London-wide · Cross-Borough Analysis

Do 6 and 9 Unit Schemes Really Get Approved More Often in London?

Abre Etteh 24 March 2026 11,694 applications with unit data, Jan 2023 onward

Retired 21 August 2026. This article is no longer part of the Perfect Scale evidence base. It is kept at its original address so that anything linking to it still resolves, and so the record of what it said stays readable. It is out of the index and out of search.

It was written to help a developer choose a unit count so as to improve the odds on a bid. That is not what this dataset is for, and the article could not be corrected into something else: the advice was the argument. The underlying observation survives — refusal rates by proposed unit count are not a smooth curve, and six units is the low point at 36.3% (n=303) against 45.4% across the development cohort — but the reason is committee thresholds and self-selection, which is a finding about how boroughs decide, not a number to design to.

Current pan-London figures are on the method, and every borough has a free dashboard. The correction history below is left in place.

Corrected August 2026, and again on 21 August. This article originally reported 35.7% and 34.3% from a 3–9 unit extract of 3,321 applications. Recomputed on the full development cohort of 11,694 decided applications, the 6-unit dip holds and the 9-unit one is much shallower than reported. The shape of the argument survives; the 9-unit figure moved 8.5 points.

That first pass corrected the headline series but left the surrounding paragraphs on the old numbers, so the article contradicted itself: it gave seven units as both 51.1% and 43.4%, and called 42.8% the lowest rate in the dataset when 36.3% is lower. The whole series now reads on one pass, and the sample sizes quoted in the limits section match the ones in the finding.

A sentence giving the 9-unit choice as “9 percentage points better” on probability of approval has gone. The gap between the 7-unit and 9-unit cohorts is 8.3 percentage points of past refusals, which describes two cohorts of decided applications, not a probability for a scheme. Full record on data releases.

The question

Developers talk about sweet spots, the unit counts that seem to get through planning more easily. Is this received wisdom, or does the data bear it out? We looked at refusal rates by proposed unit count across the 11,694 decided London small-site applications to find out where the inflection points actually sit.

The finding

The refusal rate for small sites in London is not a smooth curve. It has two distinct dips.

At 6 units, the refusal rate drops to 36.3% (n=303). At 9 units it is 42.8% (n=243). Everything between them is higher: seven units 51.1% (n=184), eight units 54.5% (n=200). The full series: one unit 45.8% (n=5,868), two 42.7% (n=2,792), three 47.6% (n=1,181), four 47.7% (n=616), five 48.2% (n=307), six 36.3% (n=303), seven 51.1% (n=184), eight 54.5% (n=200), nine 42.8% (n=243).

The band from 1 to 5 units runs between 42.7% and 48.2%. At 6 it drops to 36.3%, some 11 percentage points below the five-unit rate and the lowest of any unit count in the dataset. At 7 it jumps to 51.1%. At 8 it peaks at 54.5%. Then at 9 it falls back to 42.8%, below the whole 3-to-8 range but above the six-unit floor.

Two valleys. Two different explanations.

The 9-unit dip is the one most developers already understand. Ten units triggers affordable housing obligations in most London boroughs, typically 35% to 50% of units or a substantial financial contribution. A scheme of 9 units sits just below that threshold. Developers proposing 9 units are, by definition, optimising. They have designed to a specific constraint.

But this creates a selection effect that is important to acknowledge. The 9-unit cohort is not a random sample. These are schemes where someone (usually an experienced developer or their planning consultant) has deliberately calibrated the proposal. The schemes are likely to be better designed, better supported with technical assessments, and submitted to boroughs where the developer already understands the planning environment. The unit count does not cause the approval. The sophistication behind choosing 9 units probably does.

The 6-unit dip is less discussed and harder to explain through self-selection alone.

Six units sits below the committee referral threshold in most boroughs. Schemes of 7 or 8 units are more likely to go to planning committee rather than be decided under delegated authority. Committee decisions on small sites correlate with higher refusal rates across the dataset, as members are more risk-averse than officers on schemes at this scale. A 6-unit scheme is large enough to generate meaningful GDV but small enough to stay on an officer’s desk.

There may also be a massing explanation. On a typical London plot (150 to 250 square metres) six units often represents the natural capacity of a two-and-a-half or three storey building with sensible room sizes. It is the point where the architecture does not have to fight the site. Seven or eight units on the same footprint requires either an additional storey or tighter internal layouts, both of which attract objections.

The 8-unit peak at 54.5% is notable. It combines the worst of both effects: large enough to trigger committee in many boroughs, not large enough to justify the design investment of a 9-unit optimised scheme. It is the unit count that falls between two stools.

The implication

Consider a developer working up a scheme on a 200 square metre site in south London. The massing study suggests the site can accommodate 7 to 9 units. How should they choose?

A 7-unit scheme has a London-wide refusal rate of 51.1%. It may go to committee, depending on the borough. It sits in the dead zone, above the 6-unit sweet spot and below the 9-unit one.

A 9-unit scheme has a 42.8% refusal rate. It stays below the affordable housing threshold. In most boroughs, it remains within delegated authority. The additional two units add roughly £800,000 to £900,000 of GDV (at typical outer London values), and the 9-unit cohort has historically been refused 8.3 percentage points less often than the 7-unit one. That is a difference between two cohorts of past decisions, not a forecast for any particular scheme.

The trade-off is design quality. Fitting 9 units where 7 sit comfortably demands more from the architecture: better section design, more creative amenity provision, careful daylight geometry. The 9-unit scheme that clears planning is not the one where two extra flats are squeezed into the roof space. It is the one where 9 units were designed from the outset as the optimal capacity of the site.

This is where the self-selection effect works in the developer’s favour, if they are aware of it. The 9-unit cohort performs well because it is composed of well-prepared applications. Join that cohort. Design to 9 with the same rigour, and you inherit the statistical advantage.

The 6-unit logic operates differently. If the site naturally accommodates 6 units without strain, do not push to 7 for the additional revenue. The jump from 6 to 7 adds one unit of GDV and, historically, 14.8 percentage points of refusal: 36.3% at six units against 51.1% at seven. On a risk-adjusted basis, the seventh unit may cost more than it earns.

The limit

These are London-wide figures. Individual boroughs show different patterns; the 6-unit and 9-unit dips are not uniform everywhere. Sample sizes at specific unit counts are modest (303 at six units and 243 at nine), so borough-level breakdowns become thin. The analysis also cannot control for site size, location, or design quality. A 9-unit scheme on a large site and a 9-unit scheme on a cramped one are counted identically. And the self-selection caveat on the 9-unit finding is genuine: this is an observed correlation, not a causal claim.

Borough-level unit count profiles, committee thresholds, and affordable housing cliff effects are detailed in each of the all 33 borough reports.