Improving insights into homelessness in London with machine learning

The Mayor’s office struggles to get an accurate picture of rough sleepers in London. Each borough manages its own data, and this data is often incomplete or inconsistent. Different computer systems, teams, and providers use varied methods and formats, creating a disjointed view. What’s more, rough sleepers data can sometimes be intentionally vague, further complicating the situation. This means it’s almost impossible for the Greater London Authority and London councils to make informed decisions about where to allocate resources, to see the impact of interventions, or to understand why some people can’t get help. The London Office of Technology and Innovation (LOTI), London local government’s collaborative innovation team, wanted to see how AI could help map the movement patterns of rough sleepers.

We worked with Beam and London’s housing teams, to provide LOTI an understanding of the rough sleeper population for the first time. We developed a new insights tool that gathers and presents data from all the different data structures, boroughs and systems. Using probabilistic matching, we showcased how AI can work behind the scenes to clean and organise data. Local authorities can now see how grants are being spent, how they are supported off the streets, and how people are moving between boroughs.

So far, 40 organisations have supplied data with 151 onboarded users. The tool replaced tedious admin tasks involved in data management, and demonstrates how AI can clean and organise complex webs of fragmented data. In the future, we could connect more public sector datasets to understand broader challenges in the capital.