The question
If local authorities are refusing small-site applications for the same reasons, year after year, borough after borough, at what point does that pattern become something you can write down in advance?
The finding
We extracted and classified 13,685 individual refusal reasons from Decision Notices across all 33 London boroughs. Each reason was categorised against an 11-code taxonomy covering design, amenity, space standards, transport, heritage, daylight, policy, infrastructure, flood risk, and two catch-alls.
The distribution is concentrated:
- Design and character: 46.7%
- Neighbour amenity (overlooking, light loss, noise): 9.8%
- Space standards (unit and room sizes, outdoor space): 6.5%
- Transport and parking: 6.3%
- Daylight and sunlight: 3.2%
- Heritage and conservation: 1.9%
Combined, those six categories account for 74.4% of every refusal reason issued in London for small-site schemes. These are spatial, measurable, and repeatable criteria. They concern building height, massing, setbacks, window positions, room dimensions, parking provision, and cycle storage. They are the kind of parameters that a design code can specify.
Nearly half of every refusal reason issued in London is about character, scale, massing, or appearance. One category. One conversation, happening thousands of times.
The pattern beneath the numbers
The aggregate matters, but the borough-level consistency is what makes the case.
Westminster refuses 19% of small-site applications. Havering refuses 70.5%. Same planning system, same London Plan, same national policy framework. A 51 percentage point spread.
Within each borough, though, the refusal reasons are internally stable. Croydon cites design in 78% of its refusals, an overwhelmingly design-driven conversation. Havering's profile is different: design at 42%, neighbour amenity at 27%. Each authority has its own pattern, and that pattern holds across years and across officers.
If refusal outcomes were random or entirely dependent on individual officers, you would expect high variance within boroughs and low variance between them. The data shows the opposite. Borough-level patterns are consistent internally and distinct from each other. Each authority has a clear, if unarticulated, view of what it considers acceptable for small-site development.
The implication
The government has been pushing design codes since the Levelling Up Act. The National Model Design Code exists. Every local authority is expected to be producing them.
Almost none have done so for small sites. And small sites are where the volume is: 1 to 9 unit schemes account for the majority of applications in most boroughs. Conversions, infill plots, backland developments, end-of-terrace extensions. These schemes don't get masterplanned or sent to design review panels. They are assessed by an individual officer against policies that say things like "respond positively to local character" without specifying what that means in practice.
The result is that developers spend £15,000 to £25,000 on an application, wait 8 to 16 weeks, and discover whether their interpretation of "local character" matches the officer's. At a 43.5% aggregate failure rate across London, that is an expensive way to learn.
A design code for small sites would not need to be a 200-page masterplan document. For each area and site type, it would specify the acceptable density range, the setback and massing parameters, how to handle overlooking to the rear, what cycle and refuse storage looks like. Three or four approved precedents that demonstrate the standard.
The data to write those codes already exists. Every borough has hundreds of decided applications, with reasons given for every refusal and conditions attached to every approval. The approved schemes are the implicit code. The refused schemes mark the boundary conditions. Together, they describe what each authority will accept, whether or not the authority has articulated it.
Who benefits
Local authorities benefit because they get what they actually want. Officers currently spend weeks assessing schemes that were never going to be acceptable. A design code filters those out before submission, or reshapes them into something approvable. Officer time goes to the marginal cases that genuinely need professional judgement.
Developers benefit from certainty. A 43.5% failure rate is a system where the rules are partially hidden. Developers are asking to know what the standards are before they commit capital.
And the people who need housing benefit because homes get built. Roughly 1,600 small-site applications are refused across London each year. If even a third had been designed to a published standard and approved instead, that is over 1,000 additional homes a year from sites that already had willing developers and willing capital. No new land needed. No policy change. Less guessing.
The limit
A dataset of decided applications is the evidence base for a design code, not a replacement for the process of writing one. That requires planners, urban designers, and community input. It requires local knowledge that no spreadsheet contains.
But the dataset tells you what has been approved and refused, where, at what density, for what reasons, and with what conditions. It is the starting point that most design code exercises currently lack.
We hold that dataset for all 33 London boroughs. 12,000 applications, 13,685 classified refusal reasons, area-by-area analysis, site type breakdowns.
If your authority is spending officer time refusing applications for predictable, repeatable reasons, the question is whether that time could be better spent.