Platform Housing Group won the gold award in the innovation category at the Housing Technology 2024 Awards; in this article we take a look at the importance of these awards and what the accolade means to both Platform Housing Group and the wider housing sector.
Housing providers like Platform Housing Group are in a fantastic position to make real use of the amount of data they hold on customers. With innovative thinking and a positive, adoptive culture, there’s a real possibility of solving problems in a targeted and resource-conscious manner.
Following the Peabody report, we successfully used machine learning to identify customers who were otherwise at risk of becoming ‘silent’. We performed hundreds of tenancy health checks, stepping in and helping where, in some instances, serious self-neglect and isolation could have resulted in life-threatening problems.
AI success
We are really proud of our success in using AI over the last 18 months. This has had a very positive impact on customers and how we allocate our resources effectively, particularly in times where we need to do more for customers and under increasing financial and operational pressures; it provides a key solution to the sector doing better work for the people who need it most.
What it has also done is encourage the rest of our organisation to see what is possible with technology. To that end, we’re very proud of the culture of innovation we’ve fostered; our colleagues in our information directorate are allocated time each month to think about how we can innovate to improve our services.
As these projects grow across the rest of our organisation, it means other business areas start to see technology more as an enabler for their work. For example, our increased use of Power BI dashboards for all teams, including property defects and safeguarding, means we’re doing better work for our customers.
The most innovative steps we’ve taken in the past 18 months are where we’ve used AI, machine learning and predictive analytics to make better use of the amount of data we hold to give us real, human-focused solutions.
Silent customers
Following the Peabody report, we wanted to ask which of our 110,000 customers are more likely to be classed as silent and therefore who we should perform urgent tenancy health checks on first.
We applied the machine-learning discipline of AI to this question and experimented with both supervised and unsupervised subsets of machine learning. We trialled the ‘k-mean clustering’ algorithm based on pre-defined factors that would influence the final outcome. For example, this included asking whether they had had their gas capped, had they contacted us for a repair, had they made a complaint and if anyone else was registered in the house. To predict the likelihood of a customer becoming silent, we used a logistic regression model.
Spotting the right patterns
The benefits in our use of resources were countless. With such a large customer base, using AI helps us to direct resources in the right place and at the right time. In the case of our silent-customers work, the risk scores we produced helped us to prioritise which customers to contact first by phone and then, if necessary, in person.
Between December 2022 and June 2023, more than 500 potentially vulnerable customers were identified and contacted. Although the majority were safe and well, the contact helped us to schedule repairs and assess the different levels of support needed.
In one instance, we had a customer who hadn’t contacted us for a number of years and was found to be living in poor conditions, sleeping on the floor in a sleeping bag. We were able to open safeguarding and property condition cases and adult services began supporting the customer.
With the k-mean clustering algorithm, we used this to identify any natural groupings or patterns in the dataset to produce a risk score on customers who needed to be contacted. The logistic regression model meant that we were to quantify the possibility based on the number of times the algorithm flagged an instance that would lead to a customer not contacting us, hence putting them at risk of becoming silent.
Predicting damp and mould
In another case, we asked which properties were likely to develop damp and mould problems. For this, we implemented deep-learning AI disciplines.
First, we trialled unsupervised machine-learning models in order to get a binary outcome that would indicate the presence of damp and mould, looking at which factors in a home over a seven-day period would predict whether conditions would be right for damp and mould on the eighth day. As a refinement, we used a long short-term memory (LSTM) network so that more than one observation could be taken into account for predicting what conditions in a home would be the following day.
In the case of damp and mould predictions, our eventual choice to use a LSTM network meant that we could take a sequence of observations at a time, such as outside temperature, the condition of the home and weather forecasts.
During the initial testing of our programmes for damp and mould, we found the quality of data we were working with wasn’t good enough to produce reliable results so we’re now refining our AI product by improving the quality of the data we hold on our properties and our customers.
Essential databases & registers
The work to date has still been vital in creating essential databases and registers of our properties. We’ve made huge strides towards using AI to help tackle damp and mould, and our improved understanding of the conditions that cause it, combined with improved datasets, puts us in a great position to do more in the future.
Elsewhere, Platform Housing has innovated its systems so that our communities team can see what kind of support is needed across its geography using its ‘lower super operating model’ which analyses HACT data, local conditions, socio-economic factors and health and wellbeing assessments to target what activities it delivers.
For example, areas of high depravation now get more frequent drop-in sessions on benefit applications and access to work, and areas with high amounts of loneliness are targeted with more social-led activities such as coffee mornings and group sessions.
And that’s not all! Our development team will soon get better data on areas where regeneration is needed and our new business team can analyse potential modern methods of construction (MMC) sites.
We’ve also switched on fully-integrated cloud systems through Microsoft Dynamics 365 to link up financial services, CRM and repairs case management, giving us a 360-degree view of our services and customers for greater speed and efficiency.
Jon Cocker is the group chief information officer at Platform Housing Group. The housing provider won the gold award in the innovation category of the Housing Technology 2024 awards.
