How we measure

Every number on this site can be traced to either a sourced field in the database or a calculation performed on sourced fields. This page explains both, so that you can disagree with us precisely rather than generally.

The unit is the step, not the process

A "process" is a container. The thing we actually record is the step: one interaction with one body that must complete before something else can happen. Each step carries:

Field	What it records
`body`	Who performs it. Links to that body's page.
`type`	Application, fee, assessment, referral, notification, inspection, certification, decision, appeal, or registration.
`legal_basis`	The instrument that requires it. A step with no legal basis is either an administrative practice or a gap in our research — both worth knowing.
`statutory_clock_days`	The period the law allows, where one exists.
`typical_days`	What it takes in practice.
`wait_type`	Why it consumes calendar time.
`depends_on`	Which steps must finish first.
`confidence`	How well sourced the above are.

Why `wait_type` is the important field

Most debate about red tape founders because "it took eleven months" is not a claim anyone can act on. Eleven months of what?

So every step is classified by the reason it consumes time:

assessment — someone is genuinely evaluating the merits.
statutory — a legislated period must elapse regardless of readiness.
applicant — waiting on the applicant to supply or fix something.
third_party — waiting on an external body outside the regulator's control.
scheduling — waiting for a person or a committee to become available.
queue — waiting behind unrelated work in a backlog.

The last two are the ones this project is most interested in, because time lost to them can be recovered without lowering any standard, removing any protection, or reducing any scrutiny. Nobody is assessing anything during a scheduling wait. The file is simply waiting its turn.

That is not the same as saying such waits are waste. A panel that sits monthly may be the cheapest way to get scarce expertise into one room. It is a place to look, not a verdict.

How elapsed time is calculated

We do not add up the steps. We compute the critical path: the longest chain of dependent steps, on the assumption that steps with no dependency between them run concurrently.

This is the optimistic reading, deliberately. Where a jurisdiction does serialise independent steps, the gap between the critical path and the serial total is exactly the finding we want to surface — and reporting the optimistic figure means we can never be accused of inflating the problem.

Two consequences worth stating plainly:

Real durations are usually worse than the figures shown here. Our numbers are a floor, not an estimate of experience.
Off-critical-path delays are excluded from the wait profile. They are real but they do not lengthen the process, and counting them would overstate the prize from fixing them.

Confidence grades

No number renders on this site without a grade attached:

legislated — a section of an Act or Regulation sets this number.
published — the regulator publishes performance data showing it.
reported — a credible secondary source: an audit office, an inquiry, a parliamentary committee, substantial reporting.
practitioner — consistent testimony from people who do this for a living.
estimate — our own inference. Labelled as such wherever it appears.
unverified — a structural placeholder. Not a claim about the world.

The build script emits a warning for every duration stated without a source, and those warnings are published on the verification backlog rather than hidden in a log. A think tank that will not show you its weakest evidence is a press office.

The current seed data is almost entirely estimate grade. The step structures reflect how these pathways are actually organised, but the day counts are placeholders that exist so the analysis machinery has something to operate on. They should not be quoted. Replacing them with sourced figures is the single most valuable contribution anyone can make.

What the machine flags, and what it does not

The site automatically flags two patterns:

Steps that may not need to wait — a notification, inspection, fee or referral queued behind work by a different body.
The same body asked more than once — one authority giving multiple sign-offs of the same kind within a single process.

These are questions for a practitioner, not findings. The machine can see that a dependency exists; it cannot see whether the dependency is load-bearing. Every flag needs a person who has run the process to say whether the constraint is real. Pages label them accordingly, and we would rather publish an obvious question than a confident mistake.

What we deliberately do not measure

Whether the outcome was good. A fast process that approves a dangerous building is worse than a slow one that does not. Nothing here should be read as a claim that speed is the goal.
Cost in dollars. Fees are recorded where known but not aggregated; too much depends on project specifics for a headline number to be honest.
Anything that would let you skip a step. This is a description of process, not advice, and never a suggestion that an obligation does not apply to you.

Corrections

Every page is wrong in some way. If you have run one of these processes, the most useful thing you can tell us is which step took longer than we say, and why — because the "why" is the part that no public dataset records.

See contributing.