Lewisham identifies over 3,000 small sites with housing potential

Estimated reading time: 6 minutes

The planning challenge

Lewisham wanted to improve how it identified small sites with potential for housing delivery. In particular, it needed to:

• understand where small site opportunities exist and how many there may be
• reduce reliance on manual site identification
• support delivery on small sites alongside larger developments
• understand how existing small sites design guidance was affecting housing delivery

Small sites make an important contribution to housing delivery in the borough. However, identification of potential small sites relied heavily on officer knowledge and manual review. This made it difficult to assess the potential for housing on small sites consistently or at scale.

What they did

The pilot had two main outputs:

• an AI-based mapping tool to identify small sites with development potential
• analysis of planning application data to understand the impact of Lewisham's Small Sites Design Guide Supplementary Planning Document (SPD) on small site delivery

Developing the Small Sites AI Finder

The council worked with its supplier to develop the Small Sites AI Finder, an interactive online map designed to identify small sites with development potential.

The tool used AI and GIS (geographic information system) data to:

• analyse land parcels across the borough
• identify common types of small sites, such as infill, backland and side-street sites
• estimate the potential housing capacity of identified sites

The team trained the AI model using examples of typical small site shapes and sizes within freehold boundaries. The council used Ordnance Survey GIS mapping data to improve accuracy, although processing this data added time to the project.

Lewisham's strategic planning team and the supplier delivered the project through close collaboration. Regular review sessions helped refine how the tool presented outputs so council officers could review and understand results.

Analysing planning application data

Alongside the mapping work, the council analysed planning application data to better understand how its Small Sites Design Guide SPD was being used and its impact on decisions.

The analysis compared applications submitted before and after adoption of the guide. It reviewed application volumes, approval, refusal and withdrawal rates, and the time taken to determine applications.

Results and impact

At a glance:

• over 3,000 small sites identified with development potential
• an estimated 9,700 homes could be delivered on identified small sites
• analysis showed a small increase in approval and refusal rates following adoption of the Small Sites Design Guide SPD

Identifying small sites

The Small Sites AI Finder identified an estimated 3,017 small sites with development potential across Lewisham, with an estimated capacity of 9,747 homes.

Mapping these sites gave the council a clearer view of where opportunities were concentrated and which types of small sites were most common.

This insight could inform future policy work and engagement with landowners and developers. Lewisham's housing delivery team is now exploring the Small Sites AI Finder and considering opportunities that may be suitable for future delivery or land disposal.

Understanding the impact of the Small Sites Design Guide

The analysis compared planning applications submitted before and after adoption of the Small Sites Design Guide SPD. It reviewed application numbers, approval, refusal and withdrawal rates, and determination times.

Approval rates increased slightly, from 38.8% to 40.7%, and fewer applications were withdrawn. Refusal rates also increased, and applications took longer to determine on average.

The council highlighted that these changes may reflect closer scrutiny of planning applications following adoption of the guide, alongside wider factors such as resourcing pressures and economic conditions.

Accuracy and officer review

The pilot showed that AI-based tools can help identify sites at scale but cannot replace professional judgement.

Accuracy varied by site type, with estimates ranging from 31% to 76%. Planning officers still needed to review identified sites and cross-check them against the Small Sites Design Guide SPD. Processing Ordnance Survey GIS data improved accuracy but added time to the project.

What they learned

The council found that:

• AI tools can help identify potential sites at scale, but outputs require officer review
• the quality of training data has a significant impact on identification accuracy
• small sites vary widely in form, making automated identification challenging
• analysing planning application data alongside site identification provides useful context
• digital mapping and analysis supported better discussions across planning, housing and regeneration teams

Overall, the pilot showed that AI can support small site identification work at scale, but must currently be combined with manual officer review and other analysis to determine whether a site has potential for housing delivery.

Future plans

Lewisham Council plans to continue developing the Small Sites AI Finder to improve accuracy and functionality. The council is exploring how the tool could support targeted engagement with landowners, inform future updates to the Small Sites Design Guide SPD and help identify opportunities to increase housing delivery on small sites.

If you have feedback on this case study, you can share it using our short feedback form.

External links on this page are included to help users find relevant information. Their inclusion does not imply government endorsement of any organisation, product or service.