This case study is one of a series of ‘1 Minute Read’ studies on cost management methods published by Develin. It illustrates the power of predictive analytics to minimise future costs. One of the leading members of the G15 group of Housing Associations is engaged in a major change programme to reduce costs and grow capacity. As part of the Programme predictive analytics are pinpointing which of their General Needs Tenants are most likely to abandon their home in the near future. Even within a population of tens of thousands of tenants the numbers of abandonees is very small, perhaps 10 per month. But they matter because the costs involved are significant. It often takes a while before anyone realises that the Tenant has gone. Meanwhile arrears are mounting and the property is empty and possibly unsecured. Ongoing efforts may be underway to contact the Tenant and court action may have started. But if the Tenant can be reached before they go then support can be provided to help them turn their lives around. If they are determined to go, or have already gone, then it will at least be known about sooner, the property will lie empty for less time, and it may even be possible to recover some of what’s owed. The overall cost to the business each year is over £0.5M. Develin was asked to develop an approach to the prediction task that would help prevent as much of this cost occurring as possible.

The business of prediction

The aim was to produce a short ‘hot list’ of potential abandonees. This would allow checks to be made to confirm things one way or the other. It was also needed early enough in the life of the Tenant to help avoid costly events such as court action and eviction.

Turning the story into patterns

It was possible to see within the data for known abandonees some of the events and patterns of behaviour that pre-empted the eventual abandonment. The chart below lists some of them. These indicators, in particular signs that rent payments had stopped completely and contact with the Tenant had been lost, would therefore be central to the search.

Evidence through data

It was no surprise to find that the ideal data set that would contain all the markers and flags needed was not available. Records of key events were either sporadic or buried within larger bodies of data and scattered across multiple systems. For example, evidence of contact with the Tenant came from the Accounts Management records (payments received), the CRM system (calls received & calls made), and the Housing System (calls and visits made). Nevertheless, once all the relevant Tenant data had been assembled in one place, it became possible to compare the data for current Tenants with that for those known to have abandoned their homes. As illustrated below, Tenants could then ranked on the strength of those similarities with those at the top of the rankings becoming the ‘hot list’.

The results

To illustrate the importance of combining the different sources of data together in the way suggested the stories behind two of those on the list are presented below. “Jenny lives with her teenage son in Derby. Her working hours were reduced to 1 day per week. She was then told that she had been paid too much Housing Benefit. Her benefits would therefore be reduced until the account balanced once more. Unsurprisingly Jenny’s arrears started to mount. Her mobile number then ceased to work, gas safety contractors failed to get access, and cards through the door asking Jenny to make contact went unanswered. She was found in time and with help from the Welfare Team was able to stay. Michael lives alone in South London. His rent payments had been sporadic and occasionally he had fallen into arrears. When this happened he received reminders to pay to which he had always responded well. But he then did something not seen before. Whilst his rent account was still in credit, he phoned to check the balance. He then paid his rent for that month in full bringing his account further into the black. But he then stopped paying, and it appeared to be for good. He was found but he departed soon after. It was never known what happened.” Each of the different systems held part of each story. It was only when all the data was combined into one set of records could the whole story emerge and be capable of comparison with those relating to others.

The outcome

Out of the top 10 Tenancies on the first of the monthly hot lists, six became the subject of an investigation. Two of those received assistance that probably tipped the balance between going and staying, two were contacted at the point of going but were not persuaded to stay, and two had already departed. But, with the time to recover the properties involved shortened dramatically and the chance of recovering arrears greatly enhanced the overall cost to the business from this group alone will have fallen significantly. By repeating the exercise every month, and by widening its application e.g. to spot starter Tenancies at greatest risk of falling into arrears, a significant degree of value will have been delivered to the business.