Planning capacity is usually discussed in terms of what policy and planning documents permit. But permission on paper is only the beginning.
Between the rulebook and additional homes are several stages at which the outcome can change:
What the documents permit.
What applicants still choose to submit.
What planning authorities consent.
What that consent actually adds, measured in net homes.
Most planning analysis concentrates on the first stage. Perfect Scale is currently strongest at observing the second and third. Its work on the fourth begins with what permissions add in net homes; it does not yet measure completions across the full dataset.
The distinction matters. A policy can affect development without appearing in a refusal notice. A site can retain nominal capacity while observed decision outcomes change. A permission for two homes can add only one net home if an existing dwelling is demolished.
Perfect Scale studies those observed transitions.
It does not estimate the number of homes that would have been built if unsubmitted schemes had come forward, assign approval probabilities to individual sites, or turn observed approval rates into probability-weighted housing capacity.
The current research programme asks:
These are research questions, not positions. Some already have substantial evidence behind them; others are designs to be tested. Findings are published with their scope, limitations and the evidence needed to support them.
Transport is Croydon’s most-cited refusal category, at 28.3% of its coded reasons against a London average of 14.8%. It is the first-listed reason in 31 of 483 refusals. Design is the leading primary ground.
A published finding that pre-application advice helps in some boroughs and hurts in others turned out to be measuring the way the data was collected. The retraction, the artefact, and what would answer the question properly.
London’s ten most permissive small-sites boroughs approve above expectation for at least four different reasons. Only one of the four, a plan that states the encouragement in writing, is something another borough could adopt. Descriptive, not causal.
London’s ten most restrictive small-sites boroughs approve below expectation for more than one reason. Six can point to an adopted policy that accounts for it. Four cannot, and in one of those four the adopted text says the opposite. Descriptive, not causal.
In the middle of the table a near-zero differential is the least informative number on the page. The same headline can sit on a twenty-percentage-point gap between one site type and another, or on almost no gap at all.
London’s ten cheapest boroughs sit inside a £107,000 band and still span forty points of small-sites outcome. Price, held in this band, does not account for the spread.
Nine of London’s ten most expensive boroughs approve small sites above expectation. The tenth, Barnet, sits below it on an adopted policy that protects family housing, which makes the price gradient a tendency rather than a rule.
A typology of the 4,267 applicants behind 6,084 small-sites applications, in the 17 boroughs where the applicant is recoverable from the public record. Filing repeatedly is not associated with a higher approval rate. Descriptive, not causal.
Across 11,689 decided applications in 32 boroughs, the gap between the most permissive and most restrictive authority runs to 43.6 percentage points once caseload mix is held constant, and it tracks house prices. Descriptive, not causal.
Harrow adopted a new local plan in March 2026. For schemes of one to nine homes the change that matters is the one that did not happen. The tenure arithmetic above ten units is meaningfully different.
From 1 July 2026 the NPPF paragraph 78c twenty per cent five-year housing land supply buffer applies to seven London boroughs, with the refusal record for each.
AMR21 formally records failure against the approvals, completions and affordable supply KPIs. The same document moved the Intermediate Rent eligibility ceiling from £67,000 to £75,000.
Two regulatory dates fall inside the next eight months, and between them they pin twelve London boroughs. The second has had less coverage and probably matters more.
Camden’s affordable housing threshold is one unit. Bromley’s is eleven. The same scheme meets an eleven-fold difference in threshold depending on which side of a boundary it sits.
Of the coded refusal reasons across 33 boroughs, 61.4% cite design, amenity, space standards, daylight or transport: the categories a small-site design code can speak to.
Space standards appear in more Newham refusal notices than any other category. In a borough that refuses 61.8% of decided small-site applications, undersized units are the most common way to fail.
Brent’s 47% approval rate sits on top of a second pattern. Approved schemes lose an average of 1.2 units between application and consent, so the delivered yield is lower than the headline.
At 70.5% Havering has the highest small-site refusal rate in London. Site type, density expectation and policy context all carry part of it.
Across 5,391 decided conversion applications the London-wide approval rate is 52.5%. Among boroughs with at least 30 decided conversions the range runs from 20.0% to 92.4%.
The assumption that conservation area status lowers approval rates is widespread. Across 33 boroughs the picture is more complicated, and in some places the direction reverses.
Inner London refuses 36.5% of decided small-site applications against 49.9% in outer London. The gap is not where it is usually assumed to be.
Approved applications in Hammersmith & Fulham take substantially longer to determine than refusals. The asymmetry points to negotiation rather than delay, and it has viability consequences.
Camden conservation area sites are approved at 85.4% against 55.3% outside them, a 30 percentage point gap across 240 decided small-site applications. The direction is the opposite of the usual assumption.
Six-unit schemes are refused 36.3% of the time against a London average of 45.4%. Nine-unit schemes sit at 42.8%, better than average but not by much. The six-unit gap is the finding that survives.
Planning outcomes follow patterns that only become visible at scale. A note on what structured decision data changes, and what it does not.