Predicting Disrepair Claims in Social Housing:
Machine Learning
Identify disrepair claims trends to intervene with high risk properties before legal escalation. Ensuring organisations can allocate budgets to improvements instead of the dreaded ambulance chasers.
Introduction
On 20th March 2019 a new law came into force to make sure that rented houses and flats are ‘fit for human habitation’, which means that they are safe, healthy and free from things that could cause serious harm. While this legislation is broadly considered a positive thing to enforce for residents, it has also opened the door to a wave of opportunist law firms (ambulance chasers) amplifying the cost to resolve the core issue of disrepair.
Such firms have developed strategies that target social landlords who have dense stock in small geographies (housing estates), to backlog claims from residents who accept a small financial incentive. Once the number of cases exceed the optimal count, landlords are bombarded with multiple claims in a single day, that each have complex requirements to mitigate. Almost guaranteeing legal expenses are rubber stamped by the courts, the cost of which is often many multiples the original disrepair cost.
In this post we take a look at where the use of historic repairs and claims data can be leveraged to intervene with the root issue, and mitigate revenue being drained away from social housing into the private legal system.
Avoiding The Ambulance
Understanding the symptoms of a legal claim will often lead to the same conclusion, fix the issue to avoid the claim.
This is an overly simple statement for a complex problem, but the principal is sound. The challenge to enforce this approach is that it depends on a detailed and whole awareness of the issues that needs to be addressed. By detailed we mean the characteristics of an issue are available, and by whole we mean the various interpretations of the characteristic across the various systems and processes used to administer a property today.
All too often, the early indicators of an issue with disrepair are masked by the first handler of a complaint, either buried in a ticketing system or deescalated by the callcenter and logged in the communications record with simple text description. Even when issues arrive in the workflow, the Chinese whisper effect makes it near impossible to control the varying interpretations of the root issue. Enter Machine Learning.
By taking a whole system view, the Bimdl platform enables data to be ingested from any existing system for cross table analysis and filtering. This allows for the the most basic relationships to be established visibly and for records to be joined by their inclusion in a Bimdl Lens, displaying relevant assets in our GIS and map view for more complex processing. From here, our Machine Learning modules can be used for sentiment analysis and natural language processing to observe risk in communications that link to legal claims. Predictive training models can alert where assets are on a collision course with legal firms already sweeping through their geography.
However social landlords address mitigating the legal cost to process disrepair claims, the optimal solution will alway be preventative rather than reactive. Landlords have the data and Bimdl has the technology to ensure value remains inside the wall of a residents home, rather than in private law firms bank account.
Ingredients to begin?
The only ingredients needed to benefit from this approach are a Bimdl account and some historic data.
The strength and reliability of machine learning outcomes is heavily dependent on the data used to train the algorithm. In essence we can’t predict the future unless we have documented the past. The more training data we feed machine learning algorithm, the more they can learn.
This is exactly why social landlords are a perfect fit to apply this module of Artificial Intelligence, most have been archiving data for decades. Even where the format and structure of their data is abnormal, Bimdl’s trained machine learning models wrangle values to establish an alignment with the training models.
What are you waiting for……your next expensive disrepair claim?
Start training your ML model today
You don't need all the data to begin. Simply provide a sample of your historic data and bimdl ML will leverage our broader property data set to strengthen your model.
Submit sample dataSign up here
