Where the process blocks the policy

The Commonwealth has fifty-one stated delivery objectives that depend on administrative processes it mostly does not control: 1.2 million homes by mid-2029, hospital capacity, specialist workers in health and construction, high speed rail from Newcastle to Sydney, universal early childhood education, 82 per cent renewables — which requires transmission — aged care reform after a royal commission, defence capability it has committed to acquiring faster than it ever has, an NDIS it has committed to holding at 8 per cent growth, a migration system it has rebuilt around targeted skills, the veterans' claims system a royal commission ordered it to fix, an international education sector it is managing to a numerical target, a vocational training system it has just made a permanent commitment to, 40,000 social and affordable homes through a $10 billion fund, a disability employment system it rebuilt in November 2025, cost-of-living relief on energy bills, a student loan system it has just written down by 20 per cent, a childcare subsidy it has just uncoupled from the activity test, rent assistance it has lifted 25 per cent in two years, an employment services system it is rebuilding after an Ombudsman finding of unlawful cancellations, permanent telehealth, an aged care workforce it has spent $17.7 billion raising the wages of, faster access to new medicines, a bulk billing rate it wants at 90 per cent by 2030, an apprenticeship system it is paying up to $15,000 a head to fill, a recidivism rate it has agreed a Closing the Gap target on, and a dental system that sits almost entirely outside Medicare.

Seven of these — the NDIS, migration, veterans, international education, vocational training, disability employment and employment services — the Commonwealth controls end to end. No council, no state, no shared funding agreement in the way. They also contain the largest queue in this database, the highest refusal rate in two decades, a backlog that a royal commission ordered cleared, that was cleared, and that came back, a rule that pays more for segregated work than for open employment, and an Ombudsman finding that income support was unlawfully cancelled at scale.

This page maps each one to the specific steps, in specific bodies, that stand between the objective and its delivery — and says which tier of government can actually change each one. It is a research note, not a manifesto. Where the database contradicts the reform case, that is recorded too, and there is at least one place below where it does.

Everything here rests on process maps whose durations are almost entirely estimate grade. Read how we measure before quoting any number on this page.

The one finding that surprised us

The chain an overseas-trained specialist passes through before treating an Australian patient runs 266 days on the critical path. Its wait profile:

Zero. Not "small" — absent. The steps where Australia decides whether this specialist is safe to treat Australians, namely college assessment and registration, are not on the critical path. They run concurrently with the visa and finish first.

This matters because the reform that was actually delivered — the Expedited Specialist pathway, opened October 2024 and extended to two further specialties in January 2026 — reformed the registration step. It was a real fix to a real problem, driven by the Kruk review, and this project supports it.

But on this model it moves the completion date by roughly nothing, because registration was never the binding constraint. What binds is a visa queue governed by a ministerial direction that ranks offshore applicants below onshore ones, and a Medicare provider number check that runs last and asks a question answerable from a published dataset.

Two thirds of the elapsed time is addressable without anyone lowering a standard. See fix the steps after registration.

The general lesson generalises well beyond health: a reform aimed at the wrong step is indistinguishable from no reform, and much harder to argue with.

Housing

The objective. 1.2 million well-located homes over five years from mid-2024. Reported progress has the target slipping to around 2030, with NSW's share not expected until 2032 on some projections.

Where the time goes. On the NSW apartment process, 296 of 870 critical-path days are queue or scheduling — time in which nobody is assessing anything. The largest single component is a 70-day internal referral sequence in which individually short referrals are worked through one after another.

What is already being tried, and the awkward result. Both major fast-track pathways — the NSW Housing Delivery Authority and Victoria's Development Facilitation Program — move assessment from council to state. We mapped the HDA pathway step by step against the council pathway for the same building.

On our estimates the fast pathway runs longer end to end: 1,048 days against 870. It removes the referral sequence but adds an entry queue at the panel, a ministerial declaration, and a full environmental impact statement.

That comparison is estimate grade and should not be quoted. We publish it because it is what the model says and because it points at the real problem: there is no published, comparable, end-to-end timing data for these pathways. Declaration statistics are published — 484 proposals, about 150,000 potential homes — but conversion is where the answer is, and by mid-2026 that stood at 74 development applications lodged and 14 projects finally approved. A pathway is not fast because a minister declared you into it.

What could change, and who does it:

The Commonwealth's direct levers are the smallest of these. Its real instrument is the Accord itself: it is paying for outcomes and can attach conditions to payment. Nothing in the planning system is a Commonwealth power.

The two-sided finding on the environment track. The federal approval process runs 755 days for a subdivision that triggers it. But its largest single step, at 270 days, is the ecological survey — which happens before any application exists, and is constrained by the season in which a species can lawfully be observed. No procedural reform touches biology. The only thing that shortens it is not having to repeat a survey someone else has already done, which is bank-ecological-survey-data and is a data-publication reform, not an environmental one.

And the ANAO record cuts against the structural story: the department met statutory timeframes in 19 per cent of cases in one quarter of 2019 and 87 per cent after additional funding, with no change to the law. On that evidence a large share of federal approval delay is a staffing and queueing problem. Hiring assessors is not a reform anybody gets credit for, and it may be the highest-yield thing on this page.

Hospitals

The honest finding here is smaller than the rhetoric, and we should say so first. On the NSW hospital capital process, the critical path is 2,975 days. Construction is 1,095 of them and commissioning another 180. Addressable waits total 315 days — about 11 per cent. Building a hospital is slow mostly because building a hospital is slow.

But the composition of that 315 days is unusual. The largest single component is waiting for the annual budget round after the business case is finished and the gateway review is complete. Nobody is assessing anything during it. The project is waiting for a date in the fiscal calendar.

It is also invisible: hospital delivery is publicly measured from funding announcement, so the clock the public sees starts after this wait ends.

See decouple-design-from-budget-round, which also states the strongest counter-argument on this site — design at risk creates political commitment, and treasuries resist it for exactly that reason.

The larger hospital lever is workforce, not buildings. A commissioned hospital without staff treats nobody, which routes back to the specialist finding above, and to credentialing-verify-once — a step that repeats in full at every health service a clinician ever works at, re-verifying evidence Ahpra has already verified at source.

The Commonwealth commissioned the Scope of Practice Review (Cormack, final report October 2024, 18 recommendations) which found that virtually every primary care profession faces restrictions unrelated to their education or competence. Those restrictions are split between Commonwealth funding rules and state regulation, which is why implementation is slow. They are not yet mapped in this database and should be.

Specialist workers

Covered above for health. Two further threads:

Construction trades. The trades deficit has been forecast at 15 to 20 per cent in critical trades by late 2026. Meanwhile automatic mutual recognition — law since 2021, designed to let a licensed worker work interstate without reapplying — still excludes Queensland entirely, does not reach electrical work in NSW until October 2027, and barely touches engineers because only three jurisdictions register them.

A licensing regime that makes it administratively harder to move a qualified electrician from Brisbane to a Sydney site than to import one is working against an objective it should be neutral about. Though the honest version of that reform is licence convergence, not just recognition — see the risk section there.

Visas. The single largest recoverable delay for any imported specialist is a queue position set by ministerial direction, with no statutory clock. It is changeable by the Commonwealth alone, without legislation. It is the cheapest lever on this page.

High speed rail

We mapped the Newcastle–Sydney approvals chain in full, on both the Commonwealth and NSW tracks. The result contradicted our expectation and is published for that reason.

Critical path 2,700 days; serial total 3,190. Addressable waits: zero. The entire Commonwealth environmental track sits off the critical path, because the NSW track is longer and the two run concurrently.

So: the federal duplication on this project probably costs very little calendar time. Anyone citing this site to claim high speed rail is delayed by federal green tape is misusing it. The dominant durations are development-phase engineering, seasonal field investigation across a 194 km corridor, and land acquisition.

The case for merging the two assessments therefore rests on cost, on not asking corridor communities to submit twice about the same railway, and on fragility — a parallel track stays free only while it stays parallel, and the Commonwealth clock stops whenever further information is requested.

There is also an accountability point. The Commonwealth cannot credibly ask states to accept accredited single assessment while running a duplicate assessment of its own agency's project.

The largest lever on this project is corridor protection — reserving the alignment before land prices move — and it sits with New South Wales, not with the Commonwealth. It is not an approvals reform and this site should not pretend it is one.

Childcare

The objective. From 5 January 2026 the activity test was replaced by the 3 Day Guarantee: 72 hours per fortnight of subsidised care for all eligible families regardless of activity. That is a demand expansion, decided by the Commonwealth, effective on a fixed date.

Supply is regulated by two other tiers on no particular timetable. The NSW childcare process runs 1,000 days on the critical path, and only about 10 per cent of it is queue or scheduling. Most is genuine applicant work — finding a site with enough unencumbered outdoor space, designing, and building.

Two things do stand out.

The 150-day tail. The last three steps happen on a finished, fitted-out, empty building: service approval (90 days, statutory), then Child Care Subsidy approval (about 60 days), then open. Most of what both examine is knowable from the approved plans months earlier. The subsidy approval is a Commonwealth fitness check on an operator the state regulator has just examined, and it depends on service approval only by convention — the Commonwealth could run it concurrently on its own authority.

The reversed refusal. Childcare applications attract a consistent set of objections — traffic, parking, noise, pedestrian safety — and refusals on those grounds are reversed on appeal often enough to be a pattern. When that happens the community got no protection, the applicant paid for a court case, the council paid to defend one, and the centre opened a year late. Nobody behaved unreasonably and everyone lost. That is the argument for a code-assessable pathway, filed as an idea rather than a proposal because the thresholds are the whole argument and we do not have them.

And the necessary caveat, which is not a formality. Childcare is the one domain on this site where the direction of travel is deliberately toward more regulation, and correctly so. Since 2025 the National Law has been amended to make the child's safety, rights and best interests paramount, to create an offence of inappropriate conduct, to restrict personal digital devices in services from 27 February 2026, and to mandate national child safety training with advanced training from 31 July 2026. A national educator register, funded at $45 million, was populated by more than 96 per cent of approved services by mid-2026.

This project supports all of it and takes no position against any of it. The steps questioned above are planning and commercial steps. They are not the safety steps, and nothing here may be cited as an argument for lighter regulation of who is permitted near children. Where the two conflict — a regulator asked to go faster in the same year it is asked to be more careful — care wins.

The workforce side repeats the health finding exactly: 389 critical-path days, 285 of them queue, and zero assessment. The equivalence assessment that decides whether someone is qualified to teach under-fives runs concurrently and finishes first. The larger issue there is not a queue at all but a criterion — four years covering birth to five, which excludes teachers from comparable systems that partition the age range differently, while the shortage runs at an estimated 21,000 people. Nobody publishes how many applicants it excludes.

Energy transmission

This domain produced the best result on the site, because someone already did the reform and we can measure it.

Victoria amended the Electricity Industry Act in 2026 so that compulsory easement acquisition for transmission can proceed concurrently with the environment effects process rather than only after it — explicitly on the grounds that the sequence was delaying VNI West and the Western Renewables Link.

We modelled both pathways with every duration held identical, changing only that one dependency edge:

540 days — about eighteen months — from one edge. That is the reform, and it is the only entry on this site marked enacted, which makes it the only one that will be graded rather than argued about. If VNI West and the Western Renewables Link do not move materially, the model is wrong about something and we should say so.

It also carries a real cost that the schedule cannot see: it brings compulsory acquisition forward to a point where the route is not final, onto the party least able to bear that uncertainty. And it may not even be faster, since a landholder asked to settle against a provisional alignment has every reason to wait.

The regulatory duplication is real and is not the bottleneck. An actionable ISP project, having been found necessary by AEMO's whole-of-system cost-benefit process, then runs its own three-stage cost-benefit process. The AEMC has acknowledged this adds significantly to approval time and recommended streamlining. But merging it saves about 300 days off 3,495 — under 9 per cent — and removes two of the three formal opportunities anyone has to contest a monopoly project that households will pay for over decades. That proposal is published with its case against it, and the case against is strong.

What has actually delayed Australian transmission is landholders. VNI West slipped roughly two years on landholder engagement and assessment. The Dyer review reported in December 2023 with nine recommendations on community engagement, all accepted in principle, with a second review assessing progress in 2026. NSW and Queensland both run landholder benefit payment schemes that apply only to new projects.

A jurisdiction that copies Victoria's acquisition resequencing without the engagement reforms has taken the half that suits the proponent and left the half that suits the landholder, and should expect the resulting resistance to consume what the resequencing saved.

Aged care

This is the best-sourced process on the site and the worst result on it — and for once the numbers are the government's, not ours.

The department publishes a median of 347 days from a person's first contact with My Aged Care to their first day of Support at Home services. Our step-by-step model built independently from the published components lands at 349 days. That near-match is the first external validation this method has had, and it means the wait profile below can be taken more seriously than anything else here:

Ten per cent of a year-long wait is anyone assessing anything. The published median wait for the assessment itself is 27 days.

But most of the queue is rationing, not red tape, and this site will not pretend otherwise. About 189 days is waiting for a funded budget to be assigned against a fixed number of places. No procedural reform assigns funding that does not exist, and anyone citing this page to claim otherwise is misusing it. The department has stated an aim of three months by 2027; a Senate inquiry reports on 24 November 2026.

What process can reach is the sequence around the queue:

• The 56-day agreement window runs after assignment. A person who has waited the better part of a year — and is very often less well than when assessed — is then given eight weeks to research providers, compare them, meet them and sign. None of that needs the money to have arrived. Let it happen during the wait and the day the budget lands is the first day of service.
• Interim support is opt-in. Entry-level support exists for exactly this gap and depends on someone thinking to arrange it. Making it the default saves literally zero days, and it is on this site anyway — because a person who falls in month four doesn't get their package late, they get residential care permanently.
• The assignment queue is not published as its own number. We derived the 189 days by subtracting the agreement window from the published 245. A queue nobody publishes as a queue cannot be managed as one.

On the supply side, process is not the constraint. Building a home runs 1,581 days with only 5 per cent addressable. Reported FY25 operating EBITDA was about $5,647 per bed against an estimated $20,000–$22,000 needed to be investable, with more than 60 per cent of homes running at a deficit at above 95 per cent occupancy. Full homes losing money is not an approvals problem. The Commonwealth response is fiscal — $3bn, 5,000 extra beds a year, capital subsidies from January 2027.

And aged care supplies this project's best counter-example to its own critics. The Aged Care Act 2024 merged provider approval and accreditation into universal registration — while increasing the Commission's enforcement powers. Fewer gates and sharper teeth is a coherent design, and it is one this site should argue for more often than it does.

Defence procurement

Two processes, two different findings.

Major acquisition is slow, and the approvals are not why. The capability pathway models at 5,540 days — more than fifteen years from identified need to acceptance into service. Scheduling waits at the three government gates total 540 days, about 10 per cent. Design, build and delivery is 2,555. Anyone citing red tape as the reason Australia's capability arrives late has the wrong target; the ANAO's findings are about governance clarity and requirement-setting, not about gate duration.

What is genuinely unusual is the accountability shape. Defence is simultaneously the customer, the security regulator of its own supply chain, the author of the contract template, and the evaluator of the tender. There is no statutory clock on any step, no published elapsed time per gate, and no forum in which a supplier can contest how long something took. The Defence Delivery Agency consolidates CASG, GWEO and NSSG from 1 July 2027 — a real answer to a real finding, and the answer least likely to be measurable.

Getting into defence work is a circular dependency, written down. The SME entry process runs 810 days. A firm needs cleared people to win classified work and needs the work to justify clearing them: NV2 vetting runs about five months, DISP membership four to six more, on top of facility accreditation, ICT certification and a security officer — all incurred before any revenue.

None of the individual steps is unreasonable and this site does not ask for any of them to be shallower. But the ordering selects on balance sheet rather than capability, which works directly against the sovereign industrial capability the National Defence Strategy says it wants.

And defence has the site's cleanest deletion. Since 1 September 2024 the AUKUS licence-free environment has removed the export licence requirement for most controlled goods moving between Australia, the UK and the US. Not a faster licence — no licence. It passes the test a genuine remove should have to pass: the step's purpose was already fully served by another arrangement. It still applies in full to Japan, South Korea, Canada, New Zealand and every European partner.

It is also, again, the pattern: a real reform, delivered to a real step, that was not the binding one. A firm that cannot afford to get cleared is not helped by not needing an export licence.

The NDIS, migration, and the tribunal they both end in

These two domains are unusual here: the Commonwealth owns every step. No council, no state, no shared funding agreement. If anything on this site is squarely a federal responsibility, it is this.

The finding required no modelling. It is two published numbers next to each other.

As at 31 May 2026 the Administrative Review Tribunal held 130,983 applications, received more than 77,000 during the year, and is running an annual clearance deficit of about 30,000. Median finalisation time has risen from 45 weeks five years ago; 43 per cent of applications finish within twelve months. NDIS applicants commonly wait about a year.

And in 2025-26, 64 per cent of NDIS cases decided by the Tribunal resulted in a changed outcome — down from 73 per cent the year before.

Two thirds. The review process models at 524 days, 86 per cent of it queue and scheduling — the purest backlog in this database. The instinctive fix is capacity: more members, more sitting days. That would help, and it treats a 64 per cent change rate as a fact of nature.

It isn't one. Every one of those cases is a decision the Commonwealth made, defended for a year, and then lost or conceded. So the proposal is to empty the tribunal from the top — treat the change rate as the primary performance measure of the original decision-maker, publish it by decision type and by reason the outcome changed, and feed tribunal reasoning back into first-instance guidance.

Two serious caveats, both on that page. The sample is selected — only contested decisions reach the tribunal, and a 64 per cent change rate among them is compatible with high overall accuracy. And "reduce the overturn rate" has an ugly solution available: narrow review rights instead of improving decisions. That would register as success against the very measure this reform asks government to adopt. Any version touching review rights is not this reform, and this site would oppose it.

Migration: the tail is the story

The department publishes percentiles, and they are the most useful statistic in this database because they separate two things that are usually inseparable:

Five days against nine months — a factor of fifty-four — for the same decision, same department, same law, same form. Case complexity does not vary by fifty-four times. What varies is whether a document was missing and the department had to ask.

The median measures the decision. The gap measures the request-for-information loop, and that is where nearly all the recoverable time in Australian visa processing sits. Hence validate at lodgement and publish the RFI rate — fail in five minutes at the keyboard rather than five months in a queue.

This is also the first process on the site to use the schema's long-dormant p90_days field. Percentiles don't add, so the sum of the step tails is not the tail of the process, and the page says so.

NDIS: the clock that starts late

Access carries a 21-day Participant Service Guarantee. Reported experience is two to four months, six or more if complex. Both are true: the clock starts only when the Agency considers the application complete, and pauses whenever it asks for more. A statutory period that excludes the time the regulator controls measures the regulator's workload, not the applicant's wait — and only one of those is published.

The longest applicant-side step is 120 days gathering specialist evidence, paid for privately before any funding exists, written diagnostically by clinicians and then found incomplete for not addressing function in the scheme's terms. For permanent and unchanging conditions much of it re-proves what is already established.

On the provider side, two independently sensible commencement dates collided: mandatory registration expanded from 1 July 2026 while the first wave of five-year worker screening clearances issued in 2021 began expiring from February 2026 — new applications and renewals hitting the same eight state queues at once. Recognising one clearance nationally would make load-smoothing possible, and this site does not propose relaxing either check by a single day.

Veterans' affairs — the finding that reframes everything above

A backlog cleared is not a backlog solved, and this is the worked example.

The Royal Commission into Defence and Veteran Suicide found 41,799 unallocated DVA claims as at May 2022, and said in terms that long waits for entitlements damage veterans' mental health and in some cases contribute to suicide. It made eliminating the backlog an urgent recommendation with a 31 March 2024 deadline.

DVA cleared the last unallocated claim at the end of February 2024, ahead of the deadline. That was a real achievement by the people who did it.

As at 30 June 2026: 103,074 claims on hand, of which 32,925 unallocated. Average time to process an MRCA permanent impairment claim had risen to 364 days from 315 the year before.

Nothing here suggests bad faith. It is arithmetic. A queue empties when clearance exceeds arrival and refills when it does not. Surging staff raises clearance temporarily and does nothing to arrival, to the work each claim requires, or to the share of claims that arrive incomplete.

This should be read against most of the other pages on this site. The hospital budget wait, the tribunal, the aged care assignment queue — each could be surged down, and each would refill on the same arithmetic. Fixing the inflow means taking service records from Defence rather than making the veteran the courier, automating determinations where a Statement of Principles is plainly satisfied — in a form that can grant but never refuse, because Robodebt is the correct comparison and has to be answered first — and publishing the unallocated count monthly rather than annually.

And the reform this project would have written was already passed. The VETS Act closed two of the three veterans' compensation Acts to new claims from 1 July 2026, so every new claim is decided under a single improved MRCA. It was the royal commission's first recommendation. It is exactly the merge this database argues for.

On the department's own numbers, the queue kept growing right up to the day it commenced. Getting the structure right is necessary and it is not sufficient — and a database that only looks at structure will keep mistaking the first for the second.

Student visas

The only process here where an applicant's queue position is set by an allocation made to their institution, before they applied, which they cannot see.

Under Ministerial Direction 111, and Direction 115 for applications lodged from 14 November 2025, every active provider gets an indicative share of the national planning level — 295,000 new overseas student commencements for 2026, with no active provider allocated less than its 2025 share. Students of providers under 80 per cent of allocation are Priority 1, reported at one to four weeks. Above that, slower.

Two applicants with identical finances, English, course relevance and genuine student statements can therefore wait very different periods for reasons entirely outside themselves. That is a real policy instrument — it is how the Commonwealth manages student numbers without the legislated cap that did not pass — and this site does not argue against it.

It argues that the sequence is wrong way round. The student applies, accepts, pays a non-refundable deposit, obtains a CoE, pays a non-refundable visa charge, and only then discovers by elapsed time which queue they are in. Publishing each provider's running position saves nobody a single day and changes who is in the dark.

The decision itself shows the same tail as skilled migration — 28 days at the median, 5 months at the 90th percentile — so the RFI reform now targets it too.

Two uncomfortable numbers, recorded without an explanation attached. Refusals hit 32.5 per cent of offshore higher education applications lodged in February 2026, the highest in about two decades. And reported refusal rates vary by country of citizenship: about 65 per cent for Nepal, 51 per cent for Bangladesh, 40 per cent for India, 3 per cent for China.

This database records that range and does not explain it, because it cannot. It is consistent with genuine differences in application quality and fraud risk across cohorts; it is consistent with differential treatment; it is most likely some of both in proportions nobody outside the department can determine. What would settle it is refusal rates published by ground and by country, with review outcomes alongside.

Neither exists — and offshore student refusals generally carry no merits review right at all, so there is no change rate to measure them against. Where the NDIS has a tribunal telling it that 64 per cent of contested decisions were wrong, student visa refusals have nothing. That is a much larger transparency gap than any process finding on this page.

Childcare workforce and TAFE — one pipeline, two portfolios

These two belong together because the childcare workforce is produced by the VET system. A shortage in one is a throughput question in the other.

The placement payment stops at the wrong boundary

The Commonwealth built an instrument for placement poverty. The Commonwealth Prac Payment pays $338.60 a week in 2026, benchmarked to the single Austudy rate, to eligible students on mandatory placement. Its higher education stream covers nursing, teaching, midwifery and social work — including a Master of Teaching (Early Childhood). Its vocational stream, running since 1 July 2025, covers the Diploma of Nursing.

It does not reach a Certificate III in Early Childhood Education and Care, which requires a minimum of 160 hours of unpaid mandatory placement and is the entry qualification for most of the childcare workforce.

So in one workforce, on one government's books, the graduate route into early childhood teaching is supported during placement and the vocational route that feeds it is not. And the sector's binding constraint is teachers — at 30 June 2025, 7.4 per cent of services held a staffing waiver and almost 90 per cent related to the early childhood teacher requirement — but most early childhood teachers do not arrive as graduate entrants. They start at Certificate III and upgrade. Extending the payment saves zero days and removes a barrier at the bottom rung of the ladder.

The honest counter is on that page: there is no principled line that includes early childhood and excludes aged care, disability support or community services, all of which have unpaid mandatory placement and workforce shortages. The proposal is more expensive than it looks and should be argued at that scale.

A pay rise that depends on your employer's paperwork

The Worker Retention Payment is a 15 per cent above-award increase, extended on 17 June 2026 with $3.6 billion to 30 June 2028. It is real money reaching low-paid workers now.

But it is a grant a service opts into, conditional on a fee cap — no more than 5.8 per cent between 8 August 2026 and 7 August 2027. Two educators doing identical work next door to each other can be paid differently because one employer opted in and the other did not. Filed as an idea, not a proposal, because whether the increase could instead ride inside the subsidy rate is a legal question about Commonwealth power and the Fair Work system, not a process one — and because the reversibility that makes the current design objectionable is exactly why government chose it.

RPL: where this project's own instincts hit a wall

Recognition of prior learning is the purest form of the argument this site makes everywhere — do not assess the same thing twice. Outcome Standard 1.6 requires providers to support students with existing competence to seek recognition.

And ASQA calls inadequate RPL "an enduring and growing risk", one of its most frequent non-compliance findings, and warns about RPL mills selling qualifications for government-funded roles, shortage occupations and migration pathways.

Both are true. Recognition of real competence and sale of unearned credentials look identical from outside, and the only thing separating them is assessment rigour that is invisible in the record. Worse, funding models pay per unit delivered, so doing RPL properly earns less and costs more than teaching the course, and doing it improperly earns more than either.

Filed as an idea, and recorded at length as a corrective to this site's own method: do not assess twice is a good heuristic and not a principle, and the difference matters most where the credential is the protection.

Free TAFE: made permanent before it could be evaluated

More than 800,000 enrolments and about 258,000 completions since January 2023. The Free TAFE Act 2025 made the program permanent with over $1.6 billion to 2034-35 for at least 100,000 places a year from 2027.

Both sides of the public argument are misusing that ratio. The critics compare a three-year stock of enrolments against a flow of completions that hasn't finished arriving. The defenders' answer — the data is immature, VET takes years — is true and permanently unfalsifiable, because there is always a recent cohort still in progress.

Nobody currently knows whether Free TAFE works, including everyone confidently asserting that it does or doesn't. Reporting by starting cohort would settle it, costs no elapsed time, and uses data already collected against the Unique Student Identifier.

And a third provider-registration process

RTO registration completes a set. Three unrelated sectors, same shape:

Each starts with a long unfunded preparation phase, each asks who runs the organisation and whether it is solvent and competent, and each is followed by a second assessment before it can operate at scale. The sector-specific judgement must stay sector-specific. The fitness evidence underneath it need not — and sharing it is a safety measure as much as an efficiency one, because the fragmentation that costs providers time also costs regulators sight of each other.

Public housing

The headline wait is a survivor statistic, and that is the finding.

As at 31 March 2026 the published median wait for general applicants housed from the NSW Housing Register was 15.1 months. As at 31 December 2025 there were 67,316 households on the register — 12,094 priority, 55,222 general.

Those two numbers are usually quoted together and describe different populations. The median is computed over households that reached the front. Every household still waiting is excluded from it, and the longer someone has waited the more certain it is they are excluded from the number used to describe their wait. The published figure is a floor, and the gap grows with the backlog.

Nobody is being dishonest — Homes NSW publishes expected waiting times by area separately and claims nothing else for the median. The problem is which number travels. Publishing time-on-register for people still waiting houses nobody. It makes the scale of the rationing visible, which is the precondition for anyone being accountable for it. The allocation order is also published and is not first-come — emergency, then priority, then elderly, then wait-turn — so a general applicant's position depends on how many higher-need households arrive after them, which is unbounded.

And the counter-case to this whole site. The Housing Australia Future Fund had delivered 1,432 of its 40,000 target by May 2026 — under 4 per cent. The ANAO's July 2026 finding is procedural: applications opened January 2024, and Treasury did not finalise most governance and oversight arrangements until May 2026.

Twenty-eight months of a $10 billion program contracted before the rules for contracting were settled. This site spends most of its time arguing that steps should run earlier or concurrently — here a step ran too early, and it is published as a corrective. The test: a step can run concurrently with its successor when its output does not determine how the successor is performed. Governance determines how assessment is performed. It cannot run alongside it.

Anyone reading the Victorian easement reform should read this one too.

Disability employment

Four programs, three agencies, one person, no shared front door. A working-age person with disability deals with Services Australia for the pension, DSS through an Inclusive Employment Australia provider, the NDIA for employment supports in their plan, and DSS again through JobAccess for the Employment Assistance Fund. The chain models at 530 days.

Inclusive Employment Australia, which replaced Disability Employment Services on 1 November 2025, is a genuinely good reform — it removed time limits on support and widened eligibility to people not on Centrelink payments and to people who can only work part-time. Removing a time limit is rare in this database: for a person whose disability is permanent, a program that expires was never matched to the problem. It did not reduce the number of doors.

Then there is the rule at the end. A pension recipient may work up to 29 hours a week and keep the payment. At 30 or more, ongoing, it may be suspended. The rule does not apply to work in an Australian Disability Enterprise.

So the same person working 35 hours keeps the pension in a segregated setting and may lose it in open employment — while the Disability Royal Commission has recommended phasing segregated employment out. Most contradictions in this database are between tiers of government. This one is inside one government, between a social security rule and a disability policy commitment, expressed as a threshold in hours.

To be unmistakable: the fix is to lift open employment to the exemption, not to strip it from supported employment. Read the other way this argument would harm the people in the weakest position, and that is the opposite of what it asks.

The largest constraint in that process is none of this — it is 240 days of third_party time in job search, which is employer willingness, and no merge of Commonwealth programs touches it.

Four schemes, one shape

Social housing joins the comparison of citizen-facing scheme access, and the four together are the clearest thing this database has produced:

Four schemes, four statutes, four agencies, two tiers of government. In three of the four, the time someone spends waiting for a decision is dwarfed by the time they spend waiting for a place in a queue — and in all four, the step where somebody actually forms a judgement about the person is a minority of the elapsed time.

Most of that queue is rationing, not red tape, and every one of those pages says so. What process controls is the order, the visibility, and what happens to the person while they wait.

Energy rebates and VET student loans — the same government, two philosophies

This is the strongest argument on the site, because both halves already happened, in the same year.

In 2025 the ATO reduced every HELP and VET Student Loan debt in Australia by 20 per cent — millions of accounts, applied retrospectively to balances at 1 June 2025. No application. No form. No eligibility interview. No window to miss. No take-up gap. The legislation passed on 31 July, received assent on 3 August, and individuals did nothing at all.

In the same year, a Health Care Card holder in NSW had to reapply, annually, inside a window, for a Low Income Household Rebate worth up to $285 — for a card the Commonwealth issued, whose status Services Australia maintains, and which energy retailers already verify electronically through Centrelink Confirmation eServices with the customer's consent. The 2025-26 window closed 30 June 2026; 2026-27 opened 10 August 2026 — a 41-day scheduling gap in which nobody is assessing anything.

The usual objection to automatic payment is that government cannot know who is eligible without asking. Here it demonstrably can, the channel exists, and it is in daily commercial use. Pay concessions from the data government already holds.

The honest counter is on that page and it is uncomfortable: a rebate with an annual window has a predictable take-up rate below the eligible population, and that gap is priced into the budget. If that is the real constraint, this is a funding argument wearing an administration costume.

The edge of a good design

Commonwealth energy bill relief — $300, then $150 — reached most households as an automatic credit with no application at all. It is the best-designed delivery mechanism in this database.

Except for embedded network customers: permanent residents of caravan parks and retirement villages, and residents of many apartment buildings, who buy electricity through a landlord or body corporate. Same entitlement, but they must apply through a state portal. The reason is technical — there is no retailer to hang the credit on — and the burden lands on a group skewing older, lower-income and renting.

135 modelled days against zero for the household next door. And the model understates it, because the real cost is the households that never apply. No take-up figure for them is published, which is the first thing that should change.

Three instruments, the same rung missing from all three

VET Student Loans are a well-built scheme — provider approval, academic suitability assessment, loan caps, census-date staging. Its gates exist because VET FEE-HELP collapsed after a Senate inquiry found providers targeting vulnerable consumers, and this database does not question them: the process models at 0 per cent addressable.

The finding is the boundary. A loan cannot be used for Certificate IV or below. Put that beside the two findings from the previous section:

The qualification producing most of the childcare workforce is missing from all three, in a sector with a legislated 3 Day Guarantee and $3.6bn of wage support. Extending the determination is the narrow fix — and that page argues against itself at length, because extending loans back down toward Certificate level walks straight at the VET FEE-HELP fire, and because a fully funded place may be the right instrument rather than a debt on some of the lowest-paid work in the country.

And a remedy that requires you to find it

VET FEE-HELP debts incurred through provider misconduct can be removed. The scheme that caused the harm worked by targeting people who did not understand what they were signing. The remedy is a voluntary application.

That is a category error: an opt-in remedy selects for the capabilities whose absence caused the harm. The modelled path runs 785 days, 395 of it applicant time dominated by a step that for many people never completes at all.

Where the department has found that a provider engaged in unacceptable conduct, it knows which students enrolled with that provider in that period — from its own records. The remedy should find them.

Childcare subsidy and rent assistance

Three household payments now compare directly, and they were designed on three different philosophies:

Rent Assistance is the best-designed payment in this database. Nobody applies for it. A person claiming JobSeeker or the Age Pension is assessed for it as part of that claim. There is no window, no annual reconfirmation, and none of the take-up problem the energy rebate has. Six of its eight steps are published grade — the best-sourced process here.

The Child Care Subsidy runs on a forecast. A family estimates its combined income for a year that has not happened; that number prices twelve months of subsidy; and if it was low, a debt follows at balancing. The system knows this, which is why 5 per cent of every payment is withheld all year — a standing admission, funded by the family, that the design expects its own inputs to be wrong.

And the Commonwealth has the right number the whole time. Through Single Touch Payroll the ATO receives salary and wages each pay cycle and shares them with Services Australia, which already pre-fills that data for income support recipients — a change made expressly to stop people unknowingly accruing debts of the kind the income compliance program produced.

So the same agency, in the same year, pre-fills actual income for one payment and asks families to predict a year ahead for another. Pay the subsidy from income the government already sees.

That proposal sits next to the worst administrative failure in modern Australian government, and the page confronts it directly rather than in a footnote. Robodebt averaged annual data to infer fortnightly income and raised debts on the inference. STP is actual reported pay for actual periods, and this reform would use it prospectively to set a rate rather than retrospectively to raise a debt — it reduces debts rather than generating them. But the protections (human-reviewable estimate, easy override, no automated debt raising) are load-bearing, not decorative, and a version without them should not be built. STP also cannot see sole trader, trust or investment income, which is a real limit on exactly the variable-income households the reform is meant to help.

Two rules inside Rent Assistance

Proving the rent is the one thing a claimant must do — and without a written lease it means getting the person they pay rent to sign a form. That bites hardest where housing is least secure: lodgers, people not named on the lease, people in informal arrangements. It requires asking a favour of the person with power over your housing while disclosing that you receive income support.

The sharers rule pays a single sharer $143.60 a fortnight against $215.40 for someone not sharing — a third less, turning on whether you share a kitchen. The rule is coherent (sharing does lower per-person cost) and it taxes the main adaptation to rents rising faster than a CPI-indexed payment. Filed as an idea, because where the line should sit needs cost-of-housing evidence this project doesn't have.

Where the two tiers meet

Rent paid to a state housing authority does not attract Rent Assistance. So a household on the NSW Housing Register receives it for every month it waits — and stops the day a state dwelling is allocated.

Commonwealth recurrent money substitutes for state capital, indefinitely, and nobody publishes what that costs. The longer a state's list, the more the Commonwealth pays. Each program makes sense alone; together they let both tiers describe the problem as the other's. Publish the number — while noting it would arm two opposite bad arguments, and that it is the fourth zero-day transparency proposal on this site.

JobSeeker mutual obligations

This is the only process in the database where an instant, automated action imposes a detriment — and the asymmetry between how fast it happens and how long it takes to undo is the finding.

A contracted provider reports a failure. The payment stops the same day. Whether the report was right, and whether the person had a reasonable excuse — illness, transport, caring, a job interview, a phone out of credit — is considered afterwards.

Everywhere else in this database a zero-day step is good design: the automatic Rent Assistance assessment, the energy bill credit applied without an application, the ATO reducing every student debt in the country overnight. Australian administration is demonstrably capable of acting instantly on data it holds. The clearest instances of it doing so are a tax adjustment in the person's favour and a payment suspension against them.

What went wrong is now documented. The Commonwealth Ombudsman found 964 people had income support unlawfully cancelled between April 2022 and July 2024, and that both the cancellations and the remediation that followed were neither fair nor reasonable. Independent analysis by Economic Justice Australia put those potentially affected at more than 300,000 — a figure DEWR acknowledged at Senate Estimates as "in the vicinity of that".

964 is what an investigation with finite resources confirmed. 300,000 is what the scale of the system suggests.

And the two clocks:

Three and a half years to undo something that took an afternoon — and the route back runs through surviving without income, an Ombudsman investigation the individual did not initiate, departmental identification, and a compensation claim that has not yet been invited.

The timing is live. Penalty provisions resume progressively from 26 October 2026, with others in early 2027, while the work to identify and compensate people affected by the last round is still under way. Check before stopping someone's income, not after — and note the caution on that page: continuing payment during consideration is the whole reform, and a version that pauses payment while adding a review step is worse than what exists now.

There is also a cost nobody counts. The points target measures activity monthly, so applications get made to reach a number — and the cost of reading them falls on employers who are not part of the scheme, get nothing from it, and cannot opt out. It is the only case in this database of a scheme's costs landing on someone with no relationship to it at all. Filed as an idea: the mechanism is obviously real and the magnitude is entirely unmeasured.

Telehealth

The process is fine. The gate is the problem. A Medicare-rebated telehealth consultation models at four days and zero per cent addressable — once you're eligible, it works.

Eligibility requires a face-to-face visit with the practice in the preceding 12 months, or MyMedicare registration. That is a continuity of care protection and this site does not argue against its purpose: telehealth was made permanent on the understanding it would supplement a relationship with a regular doctor, not replace it with a stranger who has never examined you.

Its distributional effect is that telehealth is least available to people without a regular GP — people who move often, work irregular hours, are new to an area, or live where bulk-billing practices have closed.

The fix chosen was MyMedicare, and the take-up numbers are the finding: more than 2.6 million registered as at March 2025, about one in ten patients; about one in three Medicare card holders had heard of it; and 78.2 per cent of those who registered found it easy.

Easy, and unknown. The barrier is awareness, not friction. Which makes this the fourth instance of one pattern — an opt-in route reaching the people who needed it least, after the energy rebate, the embedded network application and the VET FEE-HELP remedy.

And the sixth instance of another. The practitioner is registered nationally by Ahpra; the medicine is regulated by the state. In a telehealth consultation the doctor, platform, patient and dispensing pharmacy can be in four jurisdictions, and published guidance admits it is complex and confusing to know whose law applies. Filed as an idea — and it is the instance where the case for harmonisation is weakest, because controlled-medicine regulation is a genuine state function and the states that tightened first did so after deaths. The narrow ask is a conflict rule, not uniformity.

Aged care workforce

The pattern this database has recorded six times is, in one place, being fixed.

Worker screening across eight jurisdictions, occupational licensing without Queensland, credentialing repeated at every hospital, working with children checks that don't travel, community housing registration without Victoria and WA, controlled-medicine prescribing under eight regimes — each a national scheme with someone outside it, imposing the coordination cost of a national system while delivering the benefit of none.

From mid-2026 aged care gets a national worker screening check aligned with the NDIS model, with ongoing monitoring and cross-sector recognition. A disability support worker moving into aged care is not screened from scratch. And continuous monitoring is stronger than the interim position — a police certificate under three years old tells you about someone on the day it was issued and nothing after.

Fewer checks and better protection at once, which is the same combination as aged care universal registration, and the best answer this site has to the objection that consolidating checks weakens them. Marked enacted on published guidance and a commencement date — whether cross-sector monitoring actually flows won't be observable for years.

The entry pathway is the childcare finding again. The Certificate III in Individual Support has mandatory unpaid placement and sits outside every Commonwealth student support instrument: no VET Student Loan (diploma and above only), no Commonwealth Prac Payment (its vocational stream covers the Diploma of Nursing), and a fee-free place only if the state listed it. Two sectors, the largest published workforce shortfalls in the country, the same rung missing from all three instruments.

The slowest process in the database, and the best result

Aged care wages were raised through the industrial system. The work value case was lodged in November 2020 on gender-based undervaluation grounds; the final nurses tranche is paid 1 August 2026. Five years and nine months — 2,185 modelled days, of which 1,335 is genuine assessment, because a contested tribunal proceeding establishing what care work is worth should take a long time.

Childcare wages were raised through a grant: the Worker Retention Payment, months rather than years.

Speed and durability traded directly against each other, and the slow instrument produced the better one. This site spends most of its time arguing for speed. A database that never noticed this would be measuring the wrong thing.

PBS listing

The assessment is already parallelised. The delay sits after it is finished.

The TGA and PBAC parallel process already lets registration and reimbursement evaluation run concurrently — a step resequenced in exactly the way this site usually recommends, long before this project existed. And the PBAC is a well-regarded committee whose rigour is the reason Australia pays what it does for medicines. Nothing here proposes touching it.

The modelled process runs 930 days, 39 per cent addressable — and all of that is scheduling:

• The cycle. Three meetings a year. A first submission carries about 60 days of averaged waiting. The cost compounds on resubmission, which for new medicines is common: a rejection in March is answered in May and considered in November. Out-of-cycle consideration for defined categories — though the risk section notes the cycle is also what lets busy clinicians read very large evidence packages properly.
• Cabinet. 300 of those 930 days. By this point the medicine is registered, found cost-effective, and the price is agreed. Listings above $20 million financial impact go to Cabinet rather than the Minister, and reported approval time has grown from 6.7 months to 10.

The threshold is not indexed, so inflation moves medicines across it without anyone deciding they should cross — a slightly larger share routed to the slowest forum each year by an arithmetic nobody voted for. Index it.

Two honest caveats on that page: it is a proposal to reduce Cabinet oversight of public spending and should be described that way; and the 10-month figure comes from advocacy, not a published departmental series. The Commonwealth publishing listing timeframes by stage is arguably the more important ask.

Medicare bulk billing

The steepest threshold in this database. From 1 November 2025 the Bulk Billing Practice Incentive Program pays an additional 12.5 per cent on every dollar of MBS benefit — to practices that bulk bill 100 per cent of eligible services.

Not most. All of them. And because the loading is calculated on the practice's whole MBS income, the single privately billed consultation that breaks 100 per cent doesn't cost 12.5 per cent of that consultation — it costs 12.5 per cent of everything. The marginal decision is priced at the value of the entire program.

A practice bulk billing 95 per cent is behaving in a way the policy should want, and receives nothing. Taper it — though the risk section concedes what that spends: "we bulk bill everyone" is a claim a patient can rely on, and a 92 per cent practice means nothing at a reception desk.

Seven jurisdictions picked the same lever and set it seven ways

This is the seventh instance of a national activity meeting divided jurisdictions, and it is different from the other six. The previous ones were absences — a scheme Queensland hadn't joined, a clearance that didn't travel. This one is the opposite: every jurisdiction acted, fast, in the same direction, and the incompatibility is a product of six governments independently doing the right thing.

After rulings brought contractor GP payments into the payroll tax base, each built a concession keyed to bulk billing:

They don't just differ in level — they differ in unit. A share of wages, a share of services, a proportion of income. A practice cannot compute one answer and apply it in two places, and a group operating across the Murray runs two calculations on the same billing behaviour.

Align the measurement basis, not the rates — verify once, decide separately. Though that page is honest that this is coordination in the hardest domain there is, and that the cleaner answer would be to resolve the contractor question rather than harmonise the exemptions from it.

Vocational apprenticeships

The money is at the start. So are the losses — but the money isn't there for the same reason people leave.

Trade attrition for the 2021 commencing cohort was 42.1 per cent, the highest in more than a decade, with 43 per cent completing within four years. Non-trade attrition hit a record 56 per cent. And it is front-loaded: 17.2 per cent of trade apprentices left within the first year.

NCVER records the reasons, and they are not administrative — dissatisfaction with pay or working conditions.

Now look at the money. The Key Apprenticeship Program offers up to $15,000 ($10,000 apprentice, $5,000 employer) for commencements from 1 January 2026, paid in instalments across four years. An apprentice in year one is on a training wage, usually the lowest-paid person in the workplace, receiving a fraction of a payment whose remainder arrives over three years — conditional on still being there.

Weight the payments to where the attrition is. Cost-neutral in aggregate for those who finish; it puts money in front of the people about to leave rather than behind them.

Two counter-arguments on that page carry real weight. Front-loading rewards starting, which is the arrangement that was gamed before. And attrition isn't automatically failure — nobody publishes destination data for non-completers, so 42 per cent contains outcomes ranging from personal disaster to a good decision. Treating it as one number to minimise is the error this site criticised in Free TAFE completion reporting.

The third instance of one gap

Three Commonwealth instruments supporting people who must be somewhere to train, three different answers, and the vocational pathways come last in all of them. Benchmark them the same way.

The honest objection is on that page and it is strong: an apprentice is an employee with a wage, super and leave; a student on placement is not, which is why the Prac Payment exists. Comparing the two allowances without that context overstates the gap. And if apprentice income is too low, the wage is the instrument — topping it up federally subsidises the employer rather than fixing the pay.

Prison reentry

Every system a person needs on release runs on documents, appointments and deadlines. The one thing custody reliably destroys is documents, appointments and deadlines.

A person leaving prison without a pre-release interview has seven days to claim a crisis payment, and must produce an acceptable photo ID in person at a service centre to lodge any claim — with no transport, often no fixed address, and frequently no phone credit. Proving identity requires documents; replacing any one document generally requires another; and custody is the circumstance most likely to have scattered all of them.

The pre-release mechanism that avoids all of this already exists — a claim lodged before release, an appointment arranged inside, identity authenticated in custody. It requires the release date to be known and within 21 days, an appointment to be arranged, and identity to be authenticated. It is optional in the model for exactly that reason, and the conditions fail for unplanned and remand releases — the people with the least support.

Sort out identity before the gate. Two cautions on that page: the in-person ID rule is a fraud control, not bureaucratic caution, so the reform must move when and where identity is established rather than whether; and it helps planned releases substantially and remand releases barely.

A boundary drawn by one subsection

Section 19(2) of the Health Insurance Act stops a Medicare benefit being paid for a service delivered by or under arrangement with a State. Its purpose is to prevent double-dipping. Applied to prisons, it puts people in custody outside Medicare and outside the PBS for the whole sentence.

So on release, medication supply, prescriber and funding source all change on the same day — for a population with far higher rates of chronic disease, mental illness and substance dependence than the general population, carrying no Medicare relationship and records that don't follow. The telehealth relationship requirement compounds it: a rebated consultation needs a face-to-face visit with a practice in the last 12 months, which someone who has been inside for two years does not have.

Filed as an idea, because the substance is a cost shift from state corrections budgets to the Commonwealth — precisely what s 19(2) exists to prevent — and this project can identify the boundary without being able to cost crossing it.

Dental care

Australia already has a working Commonwealth dental benefit. It stops at eighteen.

The Child Dental Benefits Schedule pays up to $1,158 over two years, with eligibility derived automatically from Medicare enrolment, age and the family's existing payments. Nobody applies. The dentist checks the balance at the chair and bills Medicare. It is the design this site argues for in automatic concession payment, already implemented.

On the eighteenth birthday it stops.

Same person, one birthday apart.

The adult service is means-tested, state-run, and reaches about one in five of the five million adults eligible for it. That is not a queue in the ordinary sense — it is a service sized for a fifth of its eligible population.

And it is funded eleven and a half months at a time. The current agreement runs 14 July 2025 to 30 June 2026 for $107.76 million, the latest in a series going back to 2013. A service cannot recruit permanent dentists, commission chairs or plan prevention against funding that expires within the year — so it buys locums and treats the emergency in front of it.

Which produces the finding that matters most here: waiting for dental care changes what treatment is possible, not just when it happens. Emergency dental is fast because it must be, and it is disproportionately extraction. A tooth that could have been filled comes out. Waiting for a house delays a house; this is the clearest case in the database of a queue producing a qualitatively different outcome.

Fund it on a real horizon — while noting that certainty about an inadequate amount is still an inadequate amount — and extend the benefit that works, which a Senate committee recommended in November 2023 and the PBO has costed at $1.143bn for 2.8 million seniors.

That last page says plainly what it is: a large recurrent spending proposal, not a process reform. And a benefit cap is not dental care — $1,158 covers exams, cleans and fillings, not the restorative work people who spent decades outside dental care actually need.

Family violence support

The order is nationally recognised. The evidence behind it is not.

Since 25 November 2017 a protection order made anywhere in Australia is automatically recognised and enforceable everywhere. A person who moves interstate does not re-apply and does not re-tell their story to a second court. That is one of only two instances in this database of the partial-national pattern being genuinely resolved, and it was done in the domain where the old arrangement was most dangerous.

What did not travel is the evidence. The same person, having left the same relationship, separately establishes what happened to:

• a court, for the order;
• the Leaving Violence Program, which needs evidence of violence and financial stress;
• Centrelink, for a crisis payment;
• a housing authority, for priority status;
• and an employer, for family violence leave.

Five bodies, five tests, each defensible alone, none designed with the others in view. Prove it once — with the direction of travel specified: the widest accepted evidence becomes sufficient everywhere, never the narrowest. The Leaving Violence Program deliberately accepts a service referral instead of a police report, because requiring police excludes people who cannot safely go to them. A "prove it once" rule built around the order as common currency would be worse than what exists.

This is also where the database's usual framing is wrong, and that page says so. The people in this process are not applicants navigating red tape; they are people at risk of being killed. Elapsed days are the wrong headline, and the 60-day estimate is the smallest part of what repeated disclosure costs.

The twelve-week window and twelve-month exclusion on the payment are filed as an idea, not a proposal: leaving is frequently not one datable event, the exclusion is tightest on the pattern that is most dangerous — and this project can see the shape of the rule without knowing what should replace it.

Mental health services

The same government, the same portfolio, the same year, built both of these:

Better Access takes about 81 days to a first session: a GP appointment, a Mental Health Treatment Plan, and finding a psychologist with capacity. Reported waits for a bulk-billed appointment run six to sixteen weeks; the alternative is to pay the gap against a rebate of $96.65 for a registered psychologist. Then a return visit to the GP for review before the later sessions.

The plan has a real clinical purpose — it keeps a GP in the loop and catches the physical causes that present as low mood. Procedurally it requires a person in distress to first obtain and pay for a different appointment in order to be allowed to pay for the one they came for. Direct referral for an initial course, with the plan required only if treatment continues — noting the strongest objection, which is that removing the GP removes the differential diagnosis.

The same shape, a third time

This is now the third pair of this kind in the database:

Three portfolios, the same finding: the better-designed instrument already exists inside the same government, applied to a fraction of the people it could serve. Expand the walk-in model — with the caveat printed on that page that the 278-day "saving" is a comparison and not a saving, because brief intervention and a course of therapy are different services.

And the funding horizon problem recurs. The National Mental Health and Suicide Prevention Agreement expired 30 June 2026 and was extended by one year — after the Productivity Commission reviewed it and found it not effective. So: a twelve-month extension of a framework the reviewing body has already said does not work. That is the dental problem with an extra turn, and the reform now targets both.

Legal aid

This is where the database's own unit breaks, and the break is the finding.

Legal assistance has the shortest critical path of any scheme-access process here — 130 days, against 612 for social housing and 487 for a DVA claim. That speed is not a virtue. It is what refusal looks like when you measure it in days.

Everywhere else in this database, unmet need shows up as a wait: a person joins a register, a queue, a list, and however badly the number is built, somebody is counted. In legal assistance the person asks and is told no. They join nothing. There is no waiting time because there is no wait.

The scale is known only from the sector, not from government: more than 1,000 people turned away daily, including about 400 victim-survivors; two-thirds of Victorians who seek help; a shortfall the Law Council puts at $500 million a year. Up to 70 per cent of those who do get through a community legal centre have an income under $32,000 — and they reached it after legal aid said no.

Count the turn-aways. This is now the fifth zero-day transparency proposal on the site, and that page says what the pattern means: they're easy to propose, produce nothing alone, and are collectively the strongest evidence that this project's real limit is missing data rather than missing ideas.

The means test gap is filed as an idea, because where the line sits is a funding question wearing a threshold's clothes — widening it without money just moves the refusal somewhere less visible. And unbundling is filed as an idea too: the duty lawyer model works because it's discrete-task, and the professional risk of advising on part of a matter is real rather than restrictive practice.

One counter-example worth recording. This site has criticised short funding horizons twice — dental at eleven and a half months, mental health extended by a single year after being found ineffective. Legal assistance did the opposite: the Mundy review reported before expiry, attorneys-general agreed terms in November 2024, and the National Access to Justice Partnership commenced 1 July 2025 with $3.9 billion over five years. Reviewed early, renegotiated early, settled for five years — in a portfolio with less political weight than health. It can be done.

Though a long agreement at an inadequate level is still inadequate, which is exactly the caution attached to the dental proposal.

Childcare deserts

Where process reform is beside the point, and the Commonwealth has just admitted it.

About 35 per cent of Australians — some 9 million people — live in a childcare desert (more than three children per place), and about 1.1 million live in regional and remote areas with no childcare at all. Deserts exist in every state and territory and every capital city. (A different threshold from the same research group gives about 24 per cent; the page quotes both, because the difference is definitional rather than a disagreement about the world.)

The ordinary pathway to a centre is not broken in a desert. It simply isn't used — a provider choosing between an outer suburb with demand and capacity to pay, and a thin market, chooses the first. And the Child Care Subsidy is a demand-side instrument: it gives families money to spend on places, and cannot conjure a place where no provider will operate. The 3 Day Guarantee expanded that demand from 5 January 2026 for every eligible family, including the ones with nothing to spend it on.

So the Commonwealth became a builder. Half the $1bn Building Early Education Fund — $500 million — is for the Australian Government to build, own and lease early childhood services. Almost every other Commonwealth instrument in this database pays somebody else to do something. This one has the Commonwealth owning the asset and carrying the risk the market declined to carry. That is a real departure and worth naming as one.

The remaining process finding is about allocation: the shortage is measured by a university, and the remedy is a competitive grant round. Thin markets are thin partly because the operators there are small, and small operators are least able to run a Commonwealth application — so a round designed to reach underserved areas is allocated by a mechanism that selects against the people in them. Publish the measure and make it a trigger, as Distribution Priority Area classification already is in health.

Two things that page states plainly: a triggered capital program in a desert with no available early childhood teacher produces a building, not a place; and about 12,000 places against 9 million people in classified deserts is a targeted intervention, not a solution to coverage.

Disaster recovery

Nothing here is a standing entitlement. Commonwealth disaster assistance is switched on by determination — per disaster, per local government area, per category. Categories A and B are standard; C and D apply in exceptional circumstances.

Through 2026 that machinery ran repeatedly: Victorian bushfires in January, NT flooding in February and March, a WA cyclone in March, North Queensland flooding across a long list of areas. And in several cases assistance was extended to further areas afterwards.

Extending on better damage information is right. What it means in the interval is that a household with a destroyed home in one LGA could claim and an identical household in the next one could not — and didn't know whether that would change. A local government area is an administrative shape; floods follow catchments and fires follow fuel and wind, and the two coincide by accident. Activate on impact, or at least publish the extension criteria up front.

But the honest process finding here is small, and the page says so. The recovery process models at 751 days with zero per cent addressable — the critical path runs through insurance assessment, development approval and the rebuild, all genuine work. Activation sits on the payments branch and resolves in weeks. This is a fairness argument about a boundary, not a speed argument.

The larger cost is invisible. A destroyed house is rebuilt against current controls — revised flood levels, bushfire attack requirements — often revised because of the event that destroyed it. That's correct policy. The insurance settlement was written against the old standard, and the difference falls on the household. Rebuilds get delayed, scaled down, or don't happen — and a household that takes a cash settlement and leaves generates no record anywhere. Filed as an idea, because funding the gap can subsidise rebuilding where the honest answer is retreat.

GP training places

The Commonwealth is adding 900 commencing places. Places were never the constraint.

Up to 1,500 fully-funded GP and rural generalist places commence each year, rising by 200 in 2026, 300 in 2027 and 400 from 2028.

The process map puts the narrowing before the application. A doctor choosing a hospital specialty stays employed by a health service; a doctor choosing general practice leaves that employment and has historically lost accrued leave, salary certainty and continuity of entitlements. Government's own stated aim is to make GP training conditions similar to hospital-based programs — which is an admission that they are not.

The most structurally unusual arrangement in the database

GP registrar pay and conditions are set by the NTCER — negotiated between General Practice Registrars Australia and GP Supervision Australia. The registrars' association and the supervisors' association. It is not an award, not an enterprise agreement, and not made by government — for a training program the Commonwealth funds, in a workforce it calls a national priority, while the hospital alternative is governed by a state award.

The tell is that from Semester 1 2026 registrars get paid study leave through a Commonwealth payment framework: the Commonwealth patching a condition in an instrument it is not party to. Put the terms somewhere binding — filed as an idea, because the NTCER's flexibility is why it adapts, and an award binding thousands of small practices is a much harder proposition.

That makes three instruments now recorded for fixing pay in Commonwealth-funded care work: an award variation in aged care (durable, five years nine months), a conditional grant in childcare (fast, reversible, expires 2028), and a private agreement plus incentive payments here.

And the fix is a trial

The Single Employer Model — one employer across all rotations, so entitlements accrue continuously and salary is fixed independently of billings — addresses exactly that step. It operates in five jurisdictions with two more from July 2026, is funded at $6.4 million to 2027-28, and is time-limited to 31 December 2028.

$6.4 million is roughly the cost of a few dozen training places, against 900 being added. Fund the fix, not the places — while conceding that a trial which has not fully reported shouldn't be made permanent on structural logic alone, which is the criticism this site levels at other people's confident reforms.

Water rights

A Commonwealth plan, four state registers, and no national view of who holds what.

Basin entitlements are issued under state law, held on state registers, traded under state approval, with state product definitions. A trade across a border is a transfer between two systems for one economic transaction.

This is the seventh instance of the partial-national pattern here and the odd one out: elsewhere there's a national scheme some jurisdictions sit outside. Here there is no national register at all. The Commonwealth plans the water and does not record who holds it — in a market it is itself a major buyer in through environmental recovery.

The process itself isn't slow: 96 days, 0% addressable. The case is the national view, not elapsed time, and the page argues it on that basis. It also keeps hydrology and administration apart deliberately — a trade that can't be delivered past the Barmah Choke can't be delivered, and no procedural reform changes the river.

And here is a step this database wanted added. From 1 July 2026, water market participants carry new disclosure obligations, market manipulation and insider trading are prohibited, and the ACCC enforces it — following the ACCC's 2021 inquiry and its 29 recommendations.

That is recorded as enacted and endorsed. A market where participants held non-public information about allocations, storage and government purchasing wasn't a free market; it was an opaque one, and opacity is a subsidy to whoever holds the information.

Five years from inquiry to commencement looks slow. This site has also criticised the Housing Australia Future Fund for opening a $10bn program before its governance was settled. Both can't be right unless the distinction is drawn properly: settling the rules before switching a regime on is worth the time it takes.

On recovery, the run rate: 450 GL is due by 31 December 2027 and about half had been recovered as at March 2026.

Remote housing

$4 billion over ten years for up to 2,700 houses is about 270 a year. The 300th was completed in 2026.

As at 13 August 2026 the Commonwealth and Northern Territory were reported deadlocked on the agreement, with building behind schedule, construction targets being reduced, reporting missing and funding not flowing — an agreement signed by both governments, all four land councils and Aboriginal Housing NT.

The context: 56.6 per cent of Aboriginal and Torres Strait Islander people in the NT live in overcrowded housing, rising to 74.5 per cent in very remote areas. Nationally 81.4 per cent were in appropriately sized housing in 2021, against 93.5 per cent of other Australians and a 2031 target of 88.

That makes the third time this database has found the same shape — a cumulative total published, and the rate required to arrive on time never published beside it:

Publish the run rate. That page also notes this is the sixth zero-day transparency proposal here, and what that means: across thirty-five Commonwealth objectives, the most common thing this project can identify is a number that isn't published. That's a finding about Australian public administration data, and a limit on what a database like this can contribute.

Where this project's method would have got it wrong

The process map has a 270-day step for obtaining a lease over Aboriginal land — the second-longest before construction. The standard method here (find the long step, ask if it can be shorter) points straight at it.

That method is wrong, and the page says why rather than quietly skipping it. Aboriginal land in the NT is inalienable freehold returned under the Aboriginal Land Rights (Northern Territory) Act 1976. A lease requires consultation with traditional owners and their informed consent. The duration is the duration of consent being properly obtained. Recording that as friction would describe the recognition of property rights as red tape — and this site wouldn't accept that framing about anyone else's freehold title.

So the question inverts: not "how is this shortened" but "is it resourced". Filed as an idea, and it will stay one — whether land councils are adequately resourced, and what tenure arrangements communities want, are questions for traditional owners and the organisations representing them. A process database can notice the question is being asked in the wrong shape. It cannot answer it.

Aged care assessments

The merge this site keeps proposing was done here, properly, and contracted for five years.

Since 9 December 2024 all aged care assessments have been conducted by a single workforce, replacing three arrangements: Regional Assessment Services, Aged Care Assessment Teams, and independent AN-ACC assessors. One standardised instrument — the Integrated Assessment Tool, twelve sections, eleven clinically validated tools — replaced several. The contracts run five years at $1.5 billion.

Set that against adult public dental funded eleven and a half months at a time and mental health extended by a single year after being found ineffective. Aged care assessment did both things right — consolidated the function and funded it on a horizon an organisation can hire against. That's the seventh reform here marked enacted, and none has been graded yet.

And the department started publishing properly

The inaugural Aged Care Act 2024 Wait Times Report (12 May 2026) publishes the median and the average together. For ongoing residential care: 167 days and 396 days.

That gap is the most informative thing in it. A median of six months and an average of thirteen means a long tail — half through in under six months, a substantial minority waiting far longer. After criticising the social housing median, the NDIS clock and visa medians throughout this site, this is the counter-example: publishing both is the correct answer to all of them.

What's still missing is the stock beside the flow:

• Median wait for a Support at Home package fell 344 → 297 days
• Number waiting rose 100,191 → 106,977 in three months
• Triage backlog fell 65,576 → 48,592

The median fell and the queue grew. Both are true. A system can get faster per person and further behind in aggregate, and only one of those numbers makes headlines.

That page also owns something: the home care process here was built against a published 347-day median and models at 349. The newer figure is 297. The model hasn't been rebuilt and the old number is still on that page with its source attached — which is what a database with estimate-grade durations looks like when better data arrives.

Skilled migration

The purest invisible non-service in this database.

The points-tested pathway models at 629 days — 81 per cent queue, and zero assessment days on the critical path. That's the fourth workforce-entry process here with zero assessment days on the critical path, after overseas doctors, overseas early childhood teachers and the sponsored skilled visa.

An applicant pays for a skills assessment and an English test, lodges an expression of interest, and waits. Invitations go to the highest scores until the allocation is exhausted — the 4 June 2026 round issued 10,000 invitations, with minimum scores from 65 points for some trades to 100 for some specialist occupations.

If your score is below the cut-off, nothing happens. Not a refusal — there is no decision, so there is nothing to be told, appealed or reviewed. You wait for the next round.

Compare the other invisible populations: legal assistance, where the person asks and is told no; energy rebates, where they never learn they were eligible. This is further along again — the person knows about the scheme, has paid, has lodged, and receives no answer of any kind. Tell them where they stand, using cut-offs the department already publishes.

That page notes this is the seventh zero-day transparency proposal here. The count is now itself the finding: across thirty-seven Commonwealth objectives, the most common thing this project can identify is a number that exists inside government and is not shared with the person it is about.

And the ground moves while you wait

Occupation lists are republished every program year. A person who paid for a skills assessment against an occupation, lodged, and is waiting can find it has moved — no decision, no notification. Planning levels move too: Skilled Independent rose from 16,900 to 21,090 places for 2026-27, inside a program of 185,000, with 35,500 more allocated across nine independent state lists.

Filed as an idea, because the obvious fix doesn't work: a list that grandfathers indefinitely stops being a shortage list. The narrower ask is notice — and the page admits that better notice mostly tells people to stop, which is valuable information nobody campaigns for.

Regional telecommunications

Three instruments for one market failure, and no sign anyone compared them.

The grant round is precise and slow: seven rounds, >$1 billion, up to 1,400 base stations, one nominated site at a time. A location must be nominated — by a community or council, so places with an active council and someone to write a submission are better represented. It waits for a round. Then a carrier must bid, and a funded location with no bidder isn't built. The process models at 940 days.

The obligation is universal and blunt: from 1 December 2027, Telstra, Optus and TPG would have to provide baseline outdoor voice and SMS coverage across Australia — feasible because LEO satellites now reach an ordinary handset. Nobody nominates anything.

Sequence them — stop running nomination rounds for baseline voice an obligation will require anyway, and redirect that funding to what the obligation doesn't reach. The page carries the symmetric caution: the obligation hasn't commenced, its definitions are contested, and winding back a working program on the strength of one that hasn't started is the error this site levelled at GP training.

The best-designed funding rule in the database

Under the current round, the subsidy rate is set by how much sharing the infrastructure delivers: up to 100 per cent where all three national carriers use active sharing or roaming, 50 to 75 per cent for co-location or fewer operators.

The Commonwealth isn't paying for a tower — it's paying more for a tower that carries everybody's customers. In thin markets, duplicated infrastructure is the classic waste, and this prices the externality directly rather than exhorting or mandating.

Almost every other funding rule in this database is binary and produces the cliffs I keep recording — the 100 per cent bulk billing requirement, the 30-hour DSP rule, the sharers rate. A graduated rate tied to a public benefit is the alternative shape, and it exists.

Drought support

Australia has no drought declaration, and that is the right design.

This site criticised disaster assistance for switching on by determination — per event, per LGA, per category — leaving households on the wrong side of a line with nothing. Drought is the counter-example. There is no declaration to be inside or outside of. Farm Household Allowance is assessed on circumstances, and the department states plainly that a common misconception is that it's a drought payment when it is available for any type of financial hardship.

That is better. Drought has no clear start, no clear end, no clean boundary; a declaration would be arbitrary at its edges and contested at its timing.

The cost of having no trigger is that nothing announces itself. A disaster activation is an event — announced, reported, and the announcement is the outreach. Drought support has no such moment. On the process map the two longest steps before an application are the hardship accumulating and the household learning the payment exists — about 300 days, and the second is entirely a communications failure.

Treat awareness as the delivery problem. Though that page carries an objection worth reading: rural communities have a documented reluctance to seek assistance, and some of that is a preference to be left alone rather than ignorance. A campaign assuming non-take-up is ignorance may misread a decision as an oversight.

And the four-years-in-ten cap runs on fixed decades, not rolling ones — so the reset is a calendar date rather than a recovery, in a country where drought cycles in the worst regions aren't reliably shorter than four years. Filed as an idea, because when support becomes a transfer to a business that won't recover is a judgement people who understand agriculture disagree about in good faith.

One thing worth noting approvingly: the refreshed Drought Resilience Hubs run to June 2032 — six years. Alongside aged care assessment (five) and legal assistance (five), against dental at eleven and a half months.

Social housing maintenance

The third leg, after access and delivery — and the one where the sequencing error is measured in decades rather than days.

In 2023, more than 30 per cent of NSW social housing stock did not meet the acceptable dwelling condition standard: four working facilities and no more than two major structural problems. That is a floor, not a quality benchmark, and a third of the portfolio was under it. Homes NSW received 24,592 urgent maintenance requests in 2023-24. A $1 billion Repair and Restore program is fixing 30,000 homes.

Maintenance is operating expenditure, competing annually against everything else. Replacement is capital. Deferring the first creates the second at several times the cost — and the accounting hides it, because the saving is booked this year and the bill lands in a decade, on a different budget line under a different minister. Nobody is ever recorded as having made the trade.

The repair process itself models at 63 days, 92 per cent queue and scheduling. And the operational reforms deserve credit: the call centre was brought back in-house and requests consolidated into a single Maintenance Hub — fewer organisations for a tenant to deal with, without removing any of the work.

A Commonwealth payment rule decides who should be the landlord

A public housing tenant attracts no Commonwealth Rent Assistance. A community housing tenant does, paid in full to the landlord as rent.

NSW has said the quiet part out loud: a key motivation for transfers is harnessing Commonwealth Rent Assistance funding previously unavailable to the social housing sector, and it is what pays for the tenancy and asset management community providers deliver. Legislation from October 2016 made transfers of tenancy management automatic.

So, with the earlier finding: the Commonwealth pays for the household in private rental, stops when the state houses them, and pays again if the state hands management to a community provider. The same dwelling, the same tenant, the same broken hot water system — and a maintenance revenue stream under one landlord and not the other.

Filed as an idea, because every fix is expensive or damaging: extending Rent Assistance to public housing would cost billions and land mostly with state budgets, and removing it from community housing would defund a sector that does the work well. The page is also careful not to claim any particular transfer was driven by money rather than merit.

Pharmacy scope

One profession, one national registration, eight scopes of practice.

A pharmacist is registered by Ahpra under a national law that applies identically everywhere. What they may do is set by eight state medicines regimes, and they have diverged sharply — Queensland treating a range of acute conditions under an Extended Practice Authority since 1 July 2025, NSW covering UTIs, contraception continuation and minor skin conditions, Victoria running a statewide pilot across 760+ pharmacies.

The same pharmacist, same training, same protocol, treats a UTI on one side of a border and refers the patient to a doctor on the other.

This is the eighth instance of the partial-national pattern here and the purest — there is no argument about different local conditions. A urinary tract infection is the same in Cairns and Ballarat. Recognise scope nationally — with the honest objection that divergence is how this got started: Queensland piloted, others followed, and a national scope settled in 2020 would have been narrower than what several states now permit.

And the pathway is the shortest in the database:

Which makes the funding gap conspicuous. Medicare pays for a consultation with a doctor and not for the same protocol-driven assessment by a pharmacist — so the state pays, or the patient does, and someone who could be treated in ten minutes has a financial reason to wait for a bulk-billed GP appointment instead. The rebate attaches to the profession, not the service.

The Commonwealth has now stepped over the top: a national women's health prescribing trial announced 13 March 2026 gives national access to UTI treatment and hormonal contraception from 1 January 2027. That's a good outcome, and the page notes the cost — it relieves exactly the pressure that would otherwise force health ministers to harmonise scope properly. A patch that makes the repair less likely.

Youth justice

The only process in this database whose critical path is 100 per cent queue and scheduling — and it is five days of a child in custody.

Start with the money and the population. Detention-based supervision is 66.2 per cent of national youth justice expenditure, about $1.1 billion a year. And about three-quarters of the young people in detention on an average night have not been sentenced — 72 per cent on the AIHW's June quarter 2025 figure, four in five on the Justice Reform Initiative's reading of the daily average.

Two-thirds of the money goes to detention, and most of the people in detention are waiting rather than serving.

The remand process models at five days, which is the published national median for a completed period of unsentenced detention. Every one of those days is queue or scheduling. None is a day of anyone evaluating the merits of anything.

A court cannot make a child a bed

Where accommodation is the obstacle, the options are release into an arrangement the court is not satisfied with, or remand. So a supply failure is recorded, in every statistic that follows, as a judicial decision about risk.

And the decision is usually reversed. Sixty per cent of completed remand periods end with the young person released on bail — the same facts, a few days later, with one thing changed. Between jurisdictions the figure runs from 15 per cent in the Northern Territory to 84 in Western Australia, a range too wide to be about different children.

The sequencing is backwards: the accommodation search happens after the bail decision, not before it. NSW already has the right instrument at the right point — the Bail and Accommodation Support Service, an after-hours service for police considering conditional bail, aimed at what the state itself calls unnecessary short-term remand. It is modelled as optional because it depends on the hour, the location, and whether a bed exists to offer. NSW has also already legislated that a child in out-of-home care must not be refused bail for want of a placement; Legal Aid NSW reports young people held in excess of a fortnight anyway.

Whether two decision-makers disagreeing 60 per cent of the time is a design or a fault is filed as an idea, not a proposal — the obvious fix is a statutory cap on remand, and several jurisdictions are currently moving hard in the opposite direction.

The step that decides everything takes no time at all

A young person can be warned, cautioned, referred to a conference, or charged. That decision is made by an officer, usually within hours, with no application, no fee, no queue and no published timeframe — which is why it appears in no process map, and why this project nearly missed it.

The Report on Government Services 2026 records that data is not available for reporting against the diversionary services indicator. There is no national caution rate.

What is known comes from research. In NSW, Aboriginal and Torres Strait Islander young people are taken to court in 64 per cent of cases against 48, and cautioned in 14 per cent against 28. Across NSW, SA and WA, Indigenous young people were less likely to be diverted in all three — after controlling for age, sex, offence type and prior history. That clause is what makes the finding hard to explain away.

Meanwhile Closing the Gap Target 11 — a 30 per cent cut in the Indigenous youth detention rate by 2031 — sits at 25.7 per 10,000 in 2024-25, below the 2018-19 baseline of 28.3 but well above the 20.5 of 2020-21. NSW, Queensland and the NT are each worsening against it. Publish the caution rate: the eighth zero-day proposal here, and the pattern holds — the domains where nothing has moved are the domains with no published number.

Where this project's method would get it wrong

In most Australian jurisdictions a 10-year-old can be charged. The ACT has raised the age to 14. The NT raised it and then lowered it again; Victoria scaled back its commitment.

That is the largest single determinant of who enters this process, and it is not a procedural question. There is no step to remove, merge or resequence. A database built to find friction in sequences has no method for a substantive rule about moral capacity, and the honest thing is to say so rather than to convert it into a process finding it does not fit. It is recorded here, and nowhere else on this site, for that reason.

Freight rail

Rail nationalised its regulator in 2013 and left its standards alone. Every other domain on this site did the opposite.

Since 2013, ONRSR has accredited every rail operator in the country under one law, with one six-month statutory clock. That is the half of the partial-national problem that no other domain here has solved.

The other half is untouched. Accreditation says an operator is competent to manage safety risk. It does not say whether a vehicle may run on a network — that is settled with each rail infrastructure manager, against that manager's own interface standards. Melbourne to Brisbane is not one network: the interstate main line is ARTC, and every capital's approaches are state-owned.

So the same test reports are assessed four times. The NTC states the consequence without qualification: in Australia it can take less time to build a train than it does to get it approved to run across the country.

The build is deliberately excluded — putting a two-year manufacturing programme on the path would make the approval look small, which is the modelling error this site has made before. Five hundred and forty days is the approval of a train that already exists.

This is the ninth instance of the partial-national pattern here, and the first where the regulator was nationalised and the standards were not. Automatic mutual recognition exists for people and not for trains — an electrician licensed in Victoria may work in Queensland without reapplying, while an inanimate object with published specifications may not.

What harmonisation actually costs when someone does it properly

The reform this site keeps proposing is, in rail, actually being done — on a published timetable:

A decade, to agree the process by which a train is approved. Not the technical specification. The process.

That timetable is itself the finding. Every harmonisation proposal on this site should be read with a ten-year clock attached, and this project has occasionally written them as though they were quick wins. The sequencing is also right, which is why it is slow: governance, then standards body, then mandatory process. Rail built ARISO before writing the standard. Occupational licensing never built the equivalent and works around it with mutual recognition — faster, and less durable.

And the interim measure earns its keep. The February 2026 Interoperability Management Plan requirement removes not one network-by-network approval; it obliges operators to think about compatibility at the point of change, making every later approval cheaper. Adding process to remove process is legitimate, and this database has been too quick elsewhere to treat an added obligation as a cost.

Inland Rail: the answer this project expected was wrong

Inland Rail went from $16.4bn (2020) to $31.4bn (2022) to an ACIL Allen estimate exceeding $45bn — more than three times its budget allocation, not deliverable before 2036. In May 2026 the government consolidated it to Beveridge to Parkes by end 2027, with everything north of Parkes reduced to corridor preservation.

Handed those numbers, a red tape database has an obvious hypothesis: environmental approvals. The Queensland sections have been in assessment for years — Border to Gowrie was still on a revised draft EIS out for consultation in mid-2025.

Dr Kerry Schott's independent review says otherwise. The delays were caused in large part by immature design and poor environmental impact statements that needed numerous changes and resubmissions, particularly in Queensland. An assessment that takes years because the proponent lodged three times is not a slow assessment. It is a resequencing failure, with an assessment cycle burned on each iteration.

Recorded as an idea, not a proposal, and mainly recorded as a correction. This is the third linear infrastructure case here where the approval was not the binding constraint — after high speed rail's federal assessment and the Housing Delivery Authority pathway that modelled longer than the council pathway it replaced.

Public transport funding

The Commonwealth committed $84 billion across 81 projects and $13.5 billion to the states in 2026-27. None of it can run a bus.

Commonwealth land transport money is capital money. It flows under the National Land Transport Act 2014, through the Infrastructure Investment Program, against the Federation Funding Agreement Schedule on Land Transport Infrastructure Projects 2024-2029 — and it is paid against construction milestones. Money moves when concrete is poured.

Then the state runs the service. Farebox recovery in Australian cities is roughly 20 to 25 cents in the dollar of operating cost. The other three-quarters is a state operating subsidy, appropriated annually, competing against schools and hospitals, for the life of the asset — and no Commonwealth program touches any part of it.

So the instrument selects the answer. Not because anyone decided rail beats buses, but because a proposal with a large capital component has a Commonwealth partner and a proposal that is purely more service has none.

What this page will not claim

The expected argument here is that bus frequency beats rail construction per dollar. The evidence assembled for this page does not support that, and saying so is more useful than asserting it: the standard Australian reference, BITRE's Benefits of Urban Public Transport Subsidies in Australia, runs on 1982-83 data and concluded that subsidies might be more efficient with lower frequency and lower fares.

The honest claim is narrower and does not depend on which option wins: one option cannot enter the competition at all. It cannot be funded even if it is the better answer, which means in practice it is never seriously assessed as one.

This is the third appearance of the same boundary in this database — deferred maintenance, where operating expenditure loses annually to capital and creates a larger capital bill a decade later; Rent Assistance, which stops the day a state asset houses someone; and now transport. Three unrelated domains, one line between what the Commonwealth will buy and what it will not.

The longest wait is waiting for a Budget

The Commonwealth decision chain — with the business case and the construction both deliberately excluded — is 480 days, and the largest single component is not assessment.

A proposal that clears evaluation in June is not refused. It waits for the next Budget. Same project, same merits, up to a year of difference depending on when the paperwork landed — and during that wait nobody is assessing anything and the project does not improve. Untie the commitment from the announcement, not the appropriation from the Budget. And publish the elapsed time between evaluation and commitment, which is not published now — which is why the 180 days is this project's estimate rather than a figure.

A threshold reform that every affected party asked for

Worth recording because it is rare. Infrastructure Australia now evaluates proposals seeking more than $250 million in Commonwealth funding, raised from $100 million, to focus evaluation on the projects that matter.

All states and territories advocated for it. A threshold reform, agreed by everyone subject to it, is the cleanest instance of this database's most common recommendation actually landing.

VET FEE-HELP

This site already mapped the student side. This is the provider side — and it is the clearest case in the database of a check being worth adding.

Getting an unfair debt removed was recorded earlier at 785 days. What was missing was how the debt got created.

Two gates, two questions, and for four years only one of them

Under VET FEE-HELP, clearing the first gate effectively reached the loan scheme, and the scheme was uncapped: revenue arrived per enrolment, from the Commonwealth, whether or not anybody learned anything.

Loans doubled between 2013 and 2014. No gate in the chain could respond to that, because none of them was looking at volume — each was assessing individual organisations against training standards, one at a time, on application.

Final cost: $2.2 billion to the taxpayer, including an estimated $1.2 billion issued inappropriately in 2014 and 2015 alone that will not be recovered.

A provider signing up students through brokers paid per head, who would never complete, was running a perfectly functional business. That is not a training-quality failure, and a training regulator was never going to catch it.

The fix, recorded as enacted

The VET Student Loans Act 2016 separated the two questions from 1 January 2017, grandfathering existing students to 31 December 2018:

• RTO registration became a precondition, not a qualification — a separate application to the department that carries the fiscal risk.
• Fitness assessed directly: financial viability, fit and proper person tests, and — tellingly — student completion rates and links to industry and employment. Neither is a training-standards question.
• A determined course list: diploma and above, with a smaller set approved only for specified listed and not-for-profit providers.
• Continuous conditions rather than trust in the gate: loan caps by course, payment in instalments against census days so a provider is not paid in full for a student who left in week two, and evidence the student is engaged.

The provider chain now models at 780 days — 570 for the training gate, 210 for the lending gate, only 8 per cent of it addressable.

The last point is the transferable one, and it is the same lesson as universal aged care registration: the durable control is a continuous, dull, cheap check, not a bigger entry assessment. Entry gates are assessed once, by people who cannot see the future, on evidence the applicant chose.

And this section is the counterweight to most of this site. Ninety-eight process maps here argue for removing, merging, thresholding or parallelising a step. This one records what happened when a check was missing.

The remedy recovered less than a fifth, and then closed

The two headline figures are not the same measurement — one is a bad-debt projection, the other is debt formally re-credited, and some of the $1.2 billion was never going to be repaid regardless. The gap is real but it is not simply money owed to victims and withheld.

What is not in doubt is the shape. The population was defined by not understanding what they signed; the remedy was a voluntary complaint with a deadline; and the window closed while the debts did not. A HELP debt is indexed, sits against a tax file number and reduces borrowing capacity for a mortgage. Someone who did not find out in time still has theirs.

And the capability is not in question. In 2025 the ATO reduced every HELP and VET Student Loan debt in the country by 20 per cent — millions of accounts, no application, no window, the same accounts. Where the department has already found that a provider engaged in unacceptable conduct, the students who enrolled with it in that period are identifiable from its own records. The finding has been made; what remains is applying it.

Childcare safety regulation

The national register took eighteen months. The clearance has taken eleven years. Same sector, same children, same ministers.

The 2025-26 package is the most substantial rewrite of Australian early childhood regulation since the National Quality Framework began, and this site's job here is mostly to record what was done rather than to propose anything.

The register is the template

The National Early Childhood Worker Register is the reform this database has proposed, in one form or another, ten times. It is the one that got built, and how it was built is the whole lesson:

• an existing national body (ACECQA) under an existing national law;
• inside the National Quality Agenda IT System the sector already used;
• populated from information providers were already required to hold;
• with a continuous 14-day update duty, which is what turns a snapshot into a monitoring instrument.

No new authority, no new data collection, no new platform. Every registry proposal on this site runs into the same three objections — who builds it, who pays for the data, who makes people use it — and this one answered all three by creating nothing new.

And one clause that is sharper than any of it

Section 284 of the National Law gave regulators two years from when an offence occurred to commence proceedings. Offences against children in care are characteristically discovered late — a child too young to disclose, a pattern visible only across several services. So the old rule produced an outcome nobody chose: the more successfully conduct was concealed, the more certainly it could not be prosecuted.

From 10 December 2025 the clock runs from awareness instead. No form, no fee, no queue — it changes when a power becomes available. That is precisely the category this database was built to find, and it is rare to find one already done.

The half that did not move

The clearance process models at 93 days, with zero addressable time — every day is genuine assessment or the applicant's own work.

The finding is not in the days. It is that the register is national and the working with children check is still eight separate schemes, each with its own Act, standards and disqualifying offences. National consistency was recommended by the Royal Commission in 2015. A worker crossing a border applies again, and waits about as long as the first time.

Progress is real: the Agreement to Deliver National Working with Children Check Reform was endorsed by attorneys-general in November 2025, and by mid-2026 all jurisdictions had agreed consistent disqualifying offences, risk-assessment benchmarks, and information sharing within jurisdictions. What is agreed only in principle is sharing across jurisdictions — which is the whole of mutual recognition — and the national criminal history checking capability is reported to be about three years away.

This is the tenth instance of the partial-national pattern here and the most uncomfortable, because the harm the check exists to prevent is the harm that moves between jurisdictions. Recognise the clearance — with the honest caveat on that page that the Royal Commission itself stopped short of recommending a single national check, and that mutual recognition is only as good as the adverse-information flow that does not yet exist.

Port infrastructure

This is the entry where the method fails, and the failure is the finding.

Australia's second-largest coal port has no container terminal. Handed that, this database goes looking for the step — a permit, an assessment, a queue. The approval chain models at 540 days with zero addressable time: state planning, EPBC assessment and a sea dumping permit, running largely in parallel, nothing recoverable, and nothing that explains anything.

The binding constraint is a contract.

The 2022 Act moved the payment; it did not remove it. The port's liability went and the State's survived, so a project that was commercially impossible became fiscally very expensive.

And none of it is misconduct. A state sold long-term rights and gave the buyer contractual protection for the revenue it was selling. The buyer paid for that protection. Two courts tested it. What this site actually asks for is not tearing up the deeds — it is that competitive-restriction terms in infrastructure privatisations be published and costed at the time of sale, because a 50-year restriction on a competing port's capacity is a fifty-year public policy decision that was made in a transaction document.

This is the fourth linear-infrastructure case here where the approval was not the binding constraint — after high speed rail's federal assessment, the Housing Delivery Authority pathway that modelled longer than the council pathway, and Inland Rail, where the Schott review blamed the proponent's own documents. Four times the instinct was wrong.

Ninety days means a year

One genuine process finding does come out of it. Section 19 of the Environment Protection (Sea Dumping) Act 1981 says the Minister must grant or refuse a permit within 90 days. The administering department advises that applications may take over 12 months depending on complexity.

Both are correct — the section provides for extension where the EPBC assessment provisions apply. But a proponent planning a seasonally-constrained dredging campaign with chartered vessels has a number in an Act that is four times shorter than the department's own expectation, and no published distribution of actual outcomes at all.

That is not a fast process or a slow one. It is an unplannable one. Publish the elapsed times — median and 90th percentile, capital dredging separated from maintenance, departmental time separated from applicant time. Ninth zero-day transparency proposal here, and the page is careful to say that the permit is not on the critical path and this is not why Australian ports do not expand.

Water infrastructure

The framework is sequenced the way this site argues planning should be sequenced. The failure happens before the framework starts.

The National Water Grid Investment Framework stages its funding: a feasibility or research study, then a preliminary business case, then a detailed business case, then construction — each required to support the next, at up to 50 per cent Commonwealth contribution. A detailed business case is required to support a construction proposal. That is a real gate doing the right thing in the right order, and this section should say so before it says anything else.

Now the portfolio: 61 large water projects, a total capital cost estimate of up to $10 billion, and 23 publicly available business cases.

A funding line sat in the forward estimates for a project whose business case then failed to support it. Nothing in the framework caused that — an announcement did, and the framework then did its job, late and expensively. Research on the sector has found costs understated by at least 116 per cent and benefits inflated by 56.

Two Commonwealth capital programs, opposite sequencing

This is the sharpest thing in the section, and it only appears because the transport work is already here.

Same Commonwealth, same year, two capital programs. Publish the business case at the moment of commitment — or publish the reason it is withheld. And note that cancelling Dungowan was the system working: a government that withdraws $595 million because the evidence does not support it has made the right call.

A dam is a container

The step this project would have missed. Building storage does not create water. Many Australian surface and groundwater systems were fully or over allocated by the 1980s — which is precisely what produced the Murray-Darling Cap in June 1995, the cap-and-trade market recorded in water trading, and the Sustainable Diversion Limits that took effect in 2019.

In a capped system a new dam buys reliability, not volume. That is a real benefit and a much smaller one than a gigalitre headline implies — and it is one mechanically plausible contributor to benefits being overstated by half.

The gate on a dam is not a permit. It is the hydrology, and a process database has nothing useful to say about it beyond insisting the business case says so out loud.

The whole pathway models at 1,990 days — five and a half years from feasibility study to construction start — of which only 180 days is recoverable, and that 180 is the same annual Budget wait found in transport.

Vocational placements

The Commonwealth pays the student. The state owns the ward. Nobody funds the supervisor.

On 23 July 2026 the Commonwealth Prac Payment was expanded to ten frontline allied health disciplines — clinical psychology, physiotherapy, occupational therapy, speech pathology, paramedicine, radiography, pharmacy, rehabilitation therapies, audiology and podiatry — from 1 July 2027, on top of teaching, nursing, midwifery and social work. Around 95,000 students a year, including 21,000 additional ones, at up to $338.60 per week.

Placement poverty is real and this is a direct answer to it. This site argued for exactly this kind of extension in the vocational stream.

And it will not create one additional placement.

Demand for clinical placements exceeded supply well before 2019, and burnout and attrition since have made it worse. The constraint is supervision: supervisor-to-student ratios traditionally as low as 1:1; supervisors required to maintain clinical delivery alongside teaching; and many registered nurses with no formal preparation for supporting student learning.

Affordability and scarcity are different constraints, and succeeding on the first makes the second more visible, not less.

This is the same boundary, for the fourth time. The Commonwealth buys the thing and the state runs it — transport capital, social housing maintenance, Rent Assistance, and now a graduate.

Five clearances, five issuers, none of them the university

The process models at 210 days with zero addressable time — and the entire critical path is the hepatitis B dose schedule and serology. That is biology. No reform touches it.

What is a decision is when the student is told. A student informed at enrolment has a scheduling problem; a student informed the semester before placement has a crisis. The whole reform there is a calendar reminder issued eighteen months earlier, and it is free.

The duplication is elsewhere, and it is concentrated where a single state has full control: host orientation modules do not travel, so a student on their third placement has done near-identical inductions three times, often for organisations inside the same state health system. Verify the student once — a statewide compliance record needs no national agreement at all.

And one thing designed well. Students are treated as volunteers for NDIS worker screening, so no fee applies. For a cohort this site keeps describing as carrying placement poverty, removing a fee from a mandatory clearance is the difference between a requirement and a barrier. It cost nothing and somebody thought about it.

Cyber security regulation

One intrusion. Five reports. Four regulators. Four clocks. And the guidance states explicitly that none of them satisfies any of the others.

This is the cleanest duplication finding in the database, and it is clean because the duplication is documented rather than inferred.

The OAIC's 30-day rule does not replace SOCI reporting — the regimes run in parallel and do not satisfy each other. Notifications to the ACSC are in addition to any other notification required at law. Nobody is confused about this. It is the design.

The defence of the design is real, and it argues for one lodgement

The regulators are asking genuinely different questions. SOCI asks whether an asset that matters to the nation was affected. The Privacy Act asks whether individuals face a risk of serious harm. CPS 234 asks whether depositors and policyholders were exposed. A single merged question would answer none of them properly.

But that is an argument against one question, not against one lodgement. File once, route automatically, deem compliance — and align the three separate 72-hour clocks, which currently run from three different definitions of awareness and constitute a compliance trap rather than a safeguard.

The process models at 127 days with zero addressable time — every day is a statutory clock. That is the point. The cost here is not elapsed time. It is the attention of the few people who understand the incident, during the hours they are containing it.

The twelve-hour clock, and why the obvious criticism is wrong

At hour twelve an organisation usually does not know what happened, and the people capable of finding out are the people writing the report. The obvious conclusion is that twelve hours is unreasonable.

The answer to that is good and this site accepts it. ASD is not reading these reports to assess the entity. It is looking for the pattern across entities — the same actor, the same vulnerability — so it can warn the next twenty organisations. For that purpose an early, vague, partly-wrong report has real value and a complete accurate one at day five has almost none. The twelve hours is not about this incident. It is about the next one.

Which changes the reform. Not a longer clock — a smaller first report: what was observed, what system, what indicators, who to contact, expressly not requiring scope, causation or impact. Anything an entity cannot know at hour twelve should not be a field at hour twelve.

And one enacted reform worth recording: the limited use obligation in the Cyber Security Act 2024 restricts how the National Cyber Security Coordinator may use what an entity discloses, to make organisations more willing to talk during an incident. It is not immunity and does not switch off any other obligation — which is the right design and is routinely over-read.

Rooftop solar

Australia has the highest per-capita rooftop solar uptake in the world, and more than half the time between deciding to install and being allowed to export is spent waiting for a meter.

The process models at 58 days, of which 30 — 52 per cent — are the meter. It is the only queue in the chain and it is the largest thing in it.

Everything before the meter works. Connection approval takes 3-10 business days. Installation takes a day.

The best subsidy delivery in this database

The small-scale technology certificates are assigned by the installer, at the point of sale. The householder makes no application, meets no window, lodges no claim and waits for no reimbursement. There is no take-up gap because there is nothing to take up.

Set that beside the energy rebate recorded earlier here, which requires an annual reapplication inside a window for a concession the Commonwealth already verifies electronically. Same policy family, opposite delivery, same government.

Since 1 July 2025 the same machinery carries the Cheaper Home Batteries Program — around 30 per cent off batteries of 5 to 100 kWh — and from 1 May 2026 the certificate factor tapers with capacity: 100 per cent to 14 kWh, 60 per cent to 28, 15 per cent to 50, with support adjusted as battery costs fall. A subsidy that declines as the technology gets cheaper and pays less for capacity beyond typical household need is the graduated design this site has argued for elsewhere.

Then the household waits for a party it never chose

The panels are on the roof. They are generating. The system is not permitted to export and the household is paid for nothing it sends out.

A regulated timeframe exists — 15 business days for digital meter installation since 1 February 2019 — and the Energy and Water Ombudsman NSW has recorded a customer more than 40 business days past notification who had contacted the retailer four or five times and been given no timeline at all. In South Australia, meter installation complaints rose 29 per cent in a year, solar upgrade delays the largest subcategory.

And nobody in this step is the household's counterparty. The installer has finished and left. The retailer appoints the metering coordinator, the coordinator does the work, the network authorises the result. The customer has a contract with one party and a complaint about another. Publish the performance, give the customer a date, and deem export permitted at the regulated timeframe — the saving is compliance with a rule that already exists.

One standard, five backstops

Rooftop solar has one national subsidy, one national installer accreditation and one national inverter standard. Then AEMO's requirement for emergency curtailment arrived five separate times:

Five thresholds, five dates, six years, for one physical problem. Eleventh instance of the partial-national pattern — and the entry argues against itself at length, because South Australia moved first for a reason, a national standard settled in 2020 would have been worse than CSIP-AUS became, and this arrangement produced the most successful energy deployment in the country's history. An entry arguing for tidier federalism should reckon with having picked an example that mostly worked.

What this exercise showed about method

Fifty-one objectives, sixty-eight processes, and a consistent pattern:

1. The reformed step is often not the binding step. Health registration, childhood teacher equivalence and high speed rail's federal assessment are all off the critical path. Reform effort went to visible steps rather than binding ones.
2. Both overseas workforce chains have zero assessment days on the critical path. A doctor and an early childhood teacher, different regulators, different laws, the same shape: every day is document gathering or a queue, and none of it is anyone judging whether they are any good. When a pattern reproduces across unrelated domains it is usually structural.
3. The biggest delays are frequently outside anyone's process. Seasonal survey windows, overseas document verification, land acquisition, landholder negotiation. These are real and no procedural change touches them.
4. Queue and scheduling waits are where the recoverable time is, and they are invisible in official statistics — because statistics are kept per body, and a queue between two bodies belongs to neither.
5. The tier that can act is usually not the tier that wants the outcome. The Commonwealth wants the homes, the hospitals, the workforce, the childcare places and the transmission. Councils and states own most of the steps. The childcare case is the sharpest: demand was expanded by Commonwealth decision on a fixed date, and supply is gated by a council and a state regulator.

6. A dependency edge can be worth eighteen months. The Victorian easement reform is the whole thesis of this database in one legislative amendment — nothing assessed differently, nothing approved more loosely, one step allowed to start earlier. It is also the only claim here that reality will grade.

7. The method now has one external check, and it passed. Aged care is the only domain where government publishes an end-to-end median. Our model, built independently from the published components, landed at 349 days against a published 347. That is one data point, on the process with the most published inputs, and it is the strongest reason to take the structure of the other twelve seriously while continuing to distrust their durations.

8. Sometimes the answer is money, and the honest thing is to say so. Aged care homes and hospital construction are both cases where process is under 10 per cent of the elapsed time and the binding constraint is capital. A site about red tape that could not reach that conclusion would be an advocacy organisation.
9. The biggest queue in the country is a symptom. A tribunal changing 64 per cent of the NDIS decisions it reviews is not a backlog to be cleared — it is an instrument measuring the decision upstream. The reflex is to fund the tribunal. The question is what would have to change for two thirds of those cases never to arrive.
10. Where percentiles are published, the tail is the finding. A median tells you the typical case; the 90th percentile tells you the case that wrecks a hiring plan or a project. Migration is the only domain here that publishes both, and the gap turns out to measure something specific and fixable — the cost of a missing document. Every other regulator in this database publishes at most an average, and several publish nothing.
11. Owning every step does not make it fast. The NDIS, migration, veterans and international education are the domains here with no council, no state and no shared funding agreement in the way. They also hold the largest queue in the database and the backlog that came back. Fragmentation is a real problem and this database has documented it repeatedly, but it is not the only one, and "if only one government were in charge" is not the lesson.
12. A queue is a rate, not an event — and this is the finding that most threatens the rest of this page. DVA cleared 41,799 unallocated claims to zero by February 2024 under a royal commission deadline, and had 32,925 unallocated again by June 2026. Several reforms here would produce a one-off improvement that refills on exactly the same arithmetic. Any proposal on this site should be read twice: once for whether it empties the queue, and once for whether it changes what fills it.
13. Where there is no review, there is no error rate. The NDIS has a tribunal telling it that 64 per cent of contested decisions were wrong. Offshore student visa refusals — running at 32.5 per cent — generally carry no merits review right, so nothing measures whether they are correct. Absence of appeals is not evidence of good decisions; it is absence of evidence.
14. Check who a published statistic is computed over. The social housing median describes households that were housed, not households waiting; the NDIS 21-day guarantee starts only once an application is deemed complete; a visa median hides a 90th percentile fifty-four times larger. In each case the number is accurate and answers a different question from the one being asked of it. This is the most common defect in Australian administrative data and it is almost never deliberate.
15. Concurrency has a limit, and it is where the earlier step's output governs the later one. The Housing Australia Future Fund opened applications twenty-eight months before its governance was settled. Most of this site argues for running steps in parallel; that page argues the opposite, and any parallelise proposal here should be tested against it.
16. Sometimes every government acts and that is the problem. Six jurisdictions built payroll tax concessions keyed to bulk billing, quickly and in the same direction, and set them in three different units — a share of wages, a share of services, a proportion of income — against a Commonwealth test of 100 per cent. Incompatibility is not always the product of inaction.
17. The same partial-national pattern keeps recurring. Worker screening, occupational licensing, clinical credentialing, working with children checks, community housing registration — each a national scheme with jurisdictions outside it. The problem is never the standard. It is that a national scheme minus two states imposes the coordination cost of a national system and delivers the benefit of neither.
18. Automatic delivery is a choice, not a capability question — and both answers exist in the same year. The ATO adjusted every student loan debt in the country in 2025 with no action from anyone. A Health Care Card holder reapplies annually, in a window, for an energy rebate whose eligibility fact the Commonwealth holds and retailers already verify electronically. Whenever a scheme says it must ask, check whether it already knows.
19. Opt-in selects against the people a scheme is for. The VET FEE-HELP remedy requires an application from people whose defining characteristic was not understanding what they signed. Interim aged care support is opt-in. Priority housing status requires the applicant to assemble third-party evidence. In each case the burden of activation lands on whoever has least capacity to carry it, and in each the state holds most of what it needs to move first.
20. Three payments, three philosophies, one government. Rent Assistance is assessed automatically on claiming another payment. The energy rebate needs an annual application inside a window. The Child Care Subsidy runs on a forecast the family makes of its own future income and charges a debt when it is wrong — while Single Touch Payroll shows the paying agency that family's actual wages every fortnight. Nothing distinguishes these designs except when each was built and what data was assumed available.
21. Instant is a capability, and it is used in both directions. The same administration that adjusts every student debt overnight and assesses Rent Assistance with no application also stops a JobSeeker payment the day a contracted provider reports a failure. Cancellation takes zero days; the route back to compensation models at 1,365. Whenever a process acts instantly, ask which way the instant runs.
22. Speed and durability trade against each other, and slow sometimes wins. Aged care wages went through a work value case: five years nine months, and a permanent award variation binding every employer. Childcare wages went through a grant: months, conditional on opting in, expiring in 2028. This site argues for speed almost everywhere; here the slow instrument produced the better one.

23. Where a payment ends is a policy boundary, and the boundaries interlock badly. Rent Assistance stops the day a state dwelling is allocated, so the Commonwealth pays more the longer a state's housing queue is. The DSP hours rule pays more for segregated work than open employment. Neither is a drafting error; both are two coherent programs meeting at an edge nobody owns.

24. A queue can change the outcome, not just its date. Waiting for a house delays a house. Waiting for public dental means the tooth that could have been filled is extracted, because emergency care is fast and restorative care is not. Where a queue is qualitative, the usual framing of recoverable delay understates the cost.

25. Some instruments already work and simply stop at a boundary. The Child Dental Benefits Schedule is automatic, means-derived and needs no application — and ends on an eighteenth birthday. The Commonwealth Prac Payment reaches nursing and teaching students and not Certificate III. The pre-release claim works and is conditional on foresight. In each case the reform is an eligibility line, not a new system.

26. The better-designed instrument is usually already inside the same government. A Medicare Mental Health Centre against Better Access, the child dental benefit against the adult waiting list, automatic Rent Assistance against an annual energy rebate application. Three portfolios, the same finding, and in each case nothing new has to be invented — only extended.

27. Sometimes elapsed time is the wrong measure entirely. In family violence, the cost of proving the same violence to five bodies is not the days it takes; it is being made to describe it again, to a new stranger, at the point of highest risk. A database built on critical paths has to say when its own unit does not fit.

28. A turn-away is not a queue, and nothing counts it. Every other unmet need in this database shows up as a waiting time. Legal assistance produces a closed door: the person asks, is refused, and joins no list — which is why it has the shortest critical path here. Where refusal is fast, elapsed time measures the wrong thing entirely.

29. Sometimes the process is fine and nobody wants to use it. The pathway to a childcare centre works; in a desert it simply is not used, because a demand-side subsidy cannot conjure a building in a thin market. The Commonwealth's answer was to become a builder and landlord, which is a real departure and an admission that process was never the constraint.

30. Capacity is not uptake. The Commonwealth is adding 900 GP training places while the instrument that determines whether doctors take them runs as a $6.4 million trial ending in 2028. A place nobody takes is not capacity; it is an unfilled place. Before funding more of something, check whether the binding constraint is its supply.

31. Three instruments, one problem, and none of them chosen on the merits. Pay and conditions in Commonwealth-funded care work have been fixed by an award variation (aged care, durable, nearly six years), a conditional grant (childcare, fast, expiring), and a private agreement between two membership bodies patched with Commonwealth payments (GP registrars). The instrument seems to follow the portfolio's history rather than the problem's shape.

32. A cumulative total without a run rate is unfalsifiable. 1,432 of 40,000 social homes. Half of 450 GL. The 300th of 2,700 remote houses. In each case a stock is published and the rate required to arrive on time is not, which lets critics call it failure and defenders call it early, forever. Publish the rate and the argument ends.

33. Some steps are rights, not friction, and the method has to be able to say so. Obtaining a lease over Aboriginal land is the second-longest step before construction in remote housing, and it is consent being properly obtained under land rights legislation. A method that could only ever recommend making steps shorter would describe the recognition of property rights as red tape. The right question there is whether the step is resourced.

34. Publishing the median and the mean together is the fix for most of the statistical complaints on this page. Aged care now does it: 167 days and 396 days for the same service. The gap makes the tail visible instead of hiding it, and it is the correct answer to the social housing median, the NDIS clock and the visa percentiles alike.

35. A median can improve while a queue grows, and both numbers are true. The Support at Home median fell from 344 days to 297 while the number waiting rose from 100,191 to 106,977. Faster per person, further behind in aggregate. Report either alone and you can tell any story.

36. The strongest form of invisible unmet need is silence, not refusal. A points-tested migration applicant pays, lodges, and if their score is below the cut-off receives nothing — no decision, so nothing to appeal. Legal assistance at least says no. This project has now found seven cases where the missing thing is a number government holds and does not share with the person it concerns.

37. Three instruments, one market failure, no comparison. A service the market will not supply in a thin market has been answered here three ways: the Commonwealth builds and owns it (childcare), pays a private firm to build it (mobile black spots), or obliges private firms to provide it (universal coverage). Nothing in any of them shows the choice being made by weighing the three.

38. A graduated rate tied to a public benefit is the alternative to a cliff. Mobile co-investment pays up to 100 per cent where all three carriers share the infrastructure and 50 to 75 per cent otherwise. Almost every other funding rule in this database is binary, and binary rules produce the thresholds this project keeps recording as harmful.

39. No trigger is sometimes the right design, and it has a cost. Drought has no declaration because drought has no clean boundary — which is better than disaster activation by local government area. But an activation announces itself and a standing entitlement does not, so the department's own material records farmers wrongly believing the payment is not for them.

40. Some sequencing errors are measured in decades, not days. Maintenance is operating expenditure and replacement is capital, so deferring the first creates the second at several times the cost — and the saving is booked this year while the bill lands in a decade, on another budget line under another minister. Nobody is ever recorded as having made the trade.

41. An eligibility rule can decide an organisational question nobody meant to ask. A public housing tenant attracts no Rent Assistance; a community housing tenant does. So the same dwelling carries a maintenance revenue stream under one landlord and not the other, and a Commonwealth payment rule is the loudest argument about who should manage Australian public housing.
42. The purest form of the partial-national pattern is a nationally registered profession with eight scopes. A pharmacist may treat a urinary tract infection on one side of a border and not the other. There is no argument here about local conditions — and the Commonwealth's answer, funding a national trial over the top, relieves the pressure that would otherwise force the states to fix it.
43. A step that takes no time can be the step that decides everything. This database measures days, so for ninety-odd process maps it looked where the days were. The most consequential step in youth justice — caution or charge — has no application, no fee, no queue and no published timeframe, and every day that follows exists because of how it went. A method built to find elapsed time will systematically miss the decisions that cost none.
44. A supply shortage gets recorded as a decision by someone who cannot fix it. A court cannot make a child a bed, so where accommodation is the obstacle the shortage comes out of the system as a judicial finding about risk — and 60 per cent of remand periods end in the same decision being reversed a few days later. Look for the step where an actor is asked to choose between options that a different actor failed to supply.
45. Sometimes it is the regulator that was nationalised and the standard that was not. Every other partial-national case here has a national registration and eight local scopes. Rail is the mirror image: one law, one regulator, one six-month clock since 2013 — and a vehicle still accepted network by network against four sets of interface standards. Solving either half alone leaves the duplication intact.
46. Harmonisation done properly takes about a decade, and this site should stop writing it as a quick win. Rail is the one domain with a published timetable: national priority in 2020, first mandatory interoperability requirements in force February 2026, mandatory approvals standard in 2029-2030. The sequencing — governance, then a standards body, then a mandatory process — is why it is slow, and it is correct. Every "recognise it nationally" proposal on this site should be read with that clock attached.
47. Sometimes the finding is that a check was missing, and this site should say so as loudly as it says the rest. VET FEE-HELP lending went from $25.6 million in 2009 to $2.9 billion in 2015 and cost the taxpayer $2.2 billion, because clearing a training-quality gate was effectively enough to originate uncapped Commonwealth debt. Ninety-eight process maps here argue for removing a step. This one argues that a step was needed and absent.
48. The durable control is a continuous cheap check, not a bigger entry gate. What actually fixed VET student lending was loan caps, instalments against census days, and evidence the student turned up — not a harder application form. Entry gates are assessed once, by people who cannot see the future, on evidence the applicant chose. The same conclusion came out of universal aged care registration, from the opposite direction.
49. A funding instrument that can only buy one kind of thing decides what gets bought. Commonwealth land transport money is capital money paid against construction milestones; three-quarters of the cost of running Australian urban public transport is a state operating subsidy no Commonwealth program touches. The point is not that buses beat rail — the evidence here does not show that. It is that one option cannot enter the competition, so it is never seriously assessed.
50. A remedy with a deadline closes on the people it was for. Redress for VET FEE-HELP re-credited $227.61 million across 33,383 complaints and shut on 31 December 2023, against $1.2 billion estimated to have been issued inappropriately in two years alone. A deadline is reasonable for a scheme whose beneficiaries know they are beneficiaries. This was not one — and the ATO reduced every HELP debt in the country by 20 per cent in 2025 without asking anybody for anything, so the constraint was never capability.
51. Harmonisation takes a decade unless something terrible happens. Point 46 said every "recognise it nationally" proposal here should carry a ten-year clock. Childcare built a mandated national worker register in about eighteen months. The variable that best explains the difference is not legal or technical difficulty — working with children checks, in the same sector for the same children, were recommended for national consistency in 2015 and are still eight schemes. It is whether a crisis has just made the reform unrefusable.
52. The registry proposals that succeed create nothing new. The National Early Childhood Worker Register used an existing national body, an existing national law, an IT system the sector already used, and data providers were already required to hold. Every registry recommendation on this site meets the same three objections — who builds it, who pays for the data, who makes people use it — and this one answered all three by adding only a duty to put existing data in an existing system.
53. Check when a power becomes available, not just how long it takes. A limitation period running from the date an offence occurred meant that in a sector where offences surface years late, the better the concealment the more certain the immunity. Moving the clock to run from awareness added no form, no fee and no queue. Timing rules that are not durations are invisible to a method that measures elapsed days, and this database nearly missed this one.
54. The binding constraint is sometimes not a step at all. Australia's second-largest coal port has no container terminal, its approval chain models at 540 days with zero recoverable time, and the actual obstacle is a 2013 compensation deed that two courts held lawful. Fourth linear-infrastructure case here where the approval was not the constraint — after high speed rail, the Housing Delivery Authority pathway and Inland Rail. A process database has to be able to conclude that the process is not the problem, and say so as plainly as it says the opposite.
55. A correctly sequenced framework does not help if the commitment arrives first. The National Water Grid Investment Framework requires a detailed business case before a construction proposal — and $595 million sat in the forward estimates for Dungowan Dam until the business case failed to support it. Sixty-one projects, up to $10 billion, twenty-three published business cases. Compare transport, where proposals over $250 million are evaluated before commitment and then wait a year for a Budget. Same Commonwealth, two capital programs, opposite failure modes.
56. Paying for access to a thing is not the same as creating the thing. The Commonwealth Prac Payment reaches around 95,000 students a year and addresses placement poverty. It creates no placements, and supervision — 1:1 ratios, supervisors carrying clinical delivery alongside teaching — is the binding constraint. Whenever a reform targets affordability, check whether the shortage is of money or of the thing itself, because succeeding on the first makes the second more visible rather than less.
57. Some constraints are physical and the honest move is to say so. Storage does not create water in a system that was fully allocated by the 1980s, and a hepatitis B schedule takes six months because that is how immunity works. Neither has a step to remove. A process database earns its credibility on the entries where it concludes that process is not the answer — and the useful residue is usually disclosure: say it in the business case, tell the student at enrolment.
58. The reform that needs no agreement is usually the one nobody has done. Host orientation modules do not travel between hospitals in the same state health system, so students repeat near-identical inductions. Fixing that needs one department to decide — no council of ministers, no national agreement, no legislation. This site spends most of its time on reforms requiring eight jurisdictions to concur, and should look harder for the ones requiring one.
59. The burden of a process is not always its elapsed time, and this site's main instrument cannot see the difference. Cyber incident reporting models at 127 days with zero addressable time — every day is a statutory clock, and the model is right. The actual cost is five reports to four regulators in the first seventy-two hours, written by the handful of people who could otherwise be containing the intrusion. A database built on days will score that as nothing to fix. It was the second domain to do this, after family violence.
60. Different questions justify separate regulators, not separate lodgements. SOCI asks whether a national asset was hit, the Privacy Act whether individuals face serious harm, CPS 234 whether depositors were exposed. Three real mandates — and the answer is one structured report, routed automatically, deemed to satisfy each regime whose fields were completed. Whenever this site finds duplication, the test is whether the question is duplicated or only the filing.
61. The best-delivered subsidy in this database is the one with nothing to apply for. Rooftop solar certificates are assigned by the installer and appear in the quoted price: no application, no window, no claim, no take-up gap. The same government requires an annual reapplication inside a window for an energy rebate whose eligibility it already verifies electronically. Delivery design is a choice, made separately each time, and the good version already exists inside the same portfolio.
62. Fragmentation is a cost, and it is sometimes not the binding one. Rooftop solar has one national subsidy, one national accreditation and one national inverter standard — and five separate emergency backstop regimes commencing across six years. It is also the most successful energy deployment in Australian history. South Australia moved first because it hit the problem first, and a national standard settled in 2020 would have been worse than what emerged. A site that recommends harmonisation this often should keep an example where the untidy version won.

And the limit that applies across domains: deregulation is not the default. The childcare section argues for faster planning steps and for more regulatory attention on safety in the same breath. Aged care's universal registration removed a gate and sharpened enforcement at the same time, which is the better template. And four things here ask government to do more, not less: interim aged care support, defence clearance sponsorship, rail's Interoperability Management Plan — an obligation added in February 2026 that removes no approval at all and makes every later one cheaper — and, most bluntly, the VET lending gate that did not exist until $2.2 billion had gone. A method that could only ever produce arguments for less process would not be a method.

And the finding this project most needs to act on itself: almost every number above is our own estimate. The comparisons that would settle these arguments — end-to-end elapsed time by pathway, per-gate assurance durations, college assessment timeframes, internal referral queue times — are not published by anyone in Australia. See the verification backlog and the companion transparency project.