The London Borough of Hackney (LBH) is an ambitious and forward-looking borough. The Borough, under new ICT leadership, is currently going through a major process of putting data and technology to work in order to much better support its service design and delivery.

We worked with the Data and Insight team to explore ways that the authority could gain better insight into and design better services for vulnerable residents, using their existing data, and always taking an ethical and practical approach to their work. This Discovery phase was very broadly defined, but at heart was intended to find where the opportunities might be to get a very broad impact from a small number of data products.

Working alongside Hackney’s data team, we discovered both operational needs – to anticipate and support vulnerable individuals’ needs better – and strategic needs – to support service leaders in measuring impact, designing new services, and predicting demand.

We also found several significant challenges. In homelessness, an area of particular importance given the high and rising levels of homelessness in the borough, we found a number of useful insights into patterns of experience based on benefits and social care data, in particular gaining a stronger understanding of the household composition and series of events that were most likely to precede the need for temporary accommodation. We hope that this will inform future housing service design for LBH. However, we also found that the existing data could not support reliable predictions of exactly which households were at risk of homelessness; since homelessness is still a relatively rare event, and since (to our surprise) the majority of households presenting as homeless were previously unknown to the benefits system, there was no chance to reliably pick them up before the event using the data available. Although this was frustrating in some ways, it was valuable for the team to understand the balance of risk and accuracy entailed in these predictions, rather than blindly using machine learning without understanding whether it would work.

We also worked with LBH to solidify their “playbook” on data and prediction ethics: working with data about the most vulnerable residents gives obvious opportunities to make people’s lives much better, but can also create significant risks of unwanted side-effects, even in the most well meaning modelling work.

The LBH team came out of this work with not only a better understanding of their back-office data and its potential to support new work, but also a stronger shared understanding and agreement of the right thing to do when building new data products, based on their own development as well as lessons and studies we could share from elsewhere. They are now ready to support LBH service delivery in delivering even better services to their residents.