Data and design to prevent homelessness: An ongoing case study
Across many social areas — from policing to identifying young people at risk of becoming Not in Education Employment or Training (NEET) — predictive data modelling is increasingly used to identify those at risk early and track their progress against the support they receive. This offers huge opportunities around creating targeted awareness channels, measuring impact and making the case for further investment. But a service design wraparound is essential. This blog highlights how we have been using predictive data within service design work, but also makes the case for a service design approach to data-led projects.
Uscreates have been working with data-led organisation Policy in Practice to support Luton Council to develop a data-led early intervention homelessness prevention service. This is part of a range of projects we are working on with partners, supporting a more systemic approach to tackling homelessness.
Luton Council already have a homelessness prevention service which supports clients up to 56 days (previously 28 days) before eviction, as required by the 2017 Homelessness Reduction Act. But using data to predict who might be at risk even before the 56 day pre-eviction period will help people to take more action themselves, either through self-service online tools or community-led support, meaning the council can prioritise its resources for those who are most in crisis. This is really exciting. The original Policy Lab and MHCLG project that informed the Trailblazer homelessness prevention funding scheme set out predictive data modelling as one of the referral routes into redesigned services. Some councils have been exploring this, but have found it challenging. So it is great to have the chance to work with Luton Council, which has been ambitious in putting into practice this pioneering approach.
Policy in Practice had already been commissioned to provide a predictive data tool, but could see the value that Uscreates could provide in designing the supportive service that those identified by the programme receive, and building the capability of frontline staff in using the data to continuously iterate the service.
Uscreates took a user-centred, design-led approach. We conducted staff observations, ethnographically informed interviews with customers, mapped community assets, co-designed an earlier outreach service and prototyped elements including the contact text or letter, online self-help wireframes. Policy in Practice also created the data model and over the summer Luton staff will be testing prototypes with a first cohort of customers identified as ‘at risk’.
Some of the biggest insights, challenges and immediate takeaways from this project so far are:
- Thinking through the ethics of data usage to identify individuals at risk. This includes making sure the data used has the correct consent, making clear to those identified why they were contacted, building in verification of their situation and providing choice over the service they receive. The initial touchpoint between the council and identified household was crucial. Even though it might have been more appropriate for a community-led service to contact those at early risk, legally it had to be the council that made contact as they are the data owners. We prototyped different channels and languages to explain the contact and to collect other data (for example personal resilience/support networks) which determine the type of support they received.
“As long as they’re going to help me I don’t care if they own my data! If they are willing to help, I would be willing to tell them anything they need.” — Customer interview.
- Early intervention and the role of the council. Providing an early intervention service went beyond the council’s statutory duty (which was to support those at risk 56 days before an eviction, but not before). We had to provide an earlier intervention service which also managed expectations and supported residents..
“Anything, any issue that is over a month, I will try and solve it myself. For anything that needs to be done in under a month, I’d go to the council.” — Customer interview
- Getting the buy-in of frontline staff in delivering a data-driven service. Frontline staff could have had concerns about the potential for data to automate their service, or felt ill-equipped to understand what the data was telling them about how their service was working. Rather than simply introducing the data tool to staff and training them in how to use it, we involved the staff throughout, interviewing them during the discovery period to understand how they currently used data, what their data needs were and what they thought — through their tacit knowledge — was the most important data to use in the model. We also introduced predictive data examples from other sectors and co-designed the data-led service with them.
- Staying open to insights found during the discovery period and introducing data-led solution later. We wanted to be able apply the data opportunities to the areas that Luton needed most, framing it correctly so that staff could see it as adding value to their work, rather than being ‘yet another computer system’. We also needed to understand user needs to develop a support service that would help prevent early homelessness.
At Uscreates, we have created a model for a service design ‘wraparound’ data projects. Over coming years, predictive data analysis and machine learning are going to become increasingly dominant. We need to make sure we can humanise this technology, using it in ways that are ethical, sensitive and understandable. Data is the lifeblood of services, and can help identify those who might benefit from them, track how people are using them so they can iteratively improve, and measure impact to make the case for them to scale up. Service designers need to be data designers.
