Some repairs teams hit their targets every month. Response times are green. Completion rates look strong. And yet somehow, the complaints keep coming in, the Housing Ombudsman cases keep mounting, and the team is exhausted. If the metrics say everything is fine, why doesn't it feel that way?

The answer might be that the metrics were never measuring the right thing.

Social housing repairs has always been a numbers game. Volume of jobs raised, average response time, operative utilisation. These figures live in dashboards and get presented at board meetings as evidence that the service is working. But they're abstractions. They tell you what happened, not why. And they can hide a lot.

There's a concept that applies here: fatal abstraction. It describes what happens when organisations make decisions based on models stripped of the causal context that determines whether those models apply. A private equity firm cuts R&D because the P&L says it improves margins. It does, briefly. What the model didn't account for was the slow erosion of competitive position those investments were quietly preventing. Boeing is another example. Cost-cutting looked right on a spreadsheet until it didn't.

In repairs, the equivalent looks like this. You measure completion rates, but not first-time fix. You measure response time, but not whether the operative arrived with the right parts. You measure jobs closed, but not whether the resident's problem was actually resolved. The numbers improve. The underlying service gets worse.

Innovation programmes are particularly exposed to this. They're often measured on outputs and activities, not resident outcomes. New tools adopted, workflows implemented, reports produced. None of which tells you whether a single resident is better off than they were before.

Here's where AI changes the stakes. AI is very good at taking your existing assumptions and making them sound comprehensive, well-reasoned, and complete. If the underlying model is right, that's enormously useful. If the model is wrong, AI scales the wrong answer, quickly and convincingly. As one framework puts it: garbage in, garbage out, but now at scale and beautifully formatted.

Placing an AI tool inside a repairs service that's measuring the wrong things doesn't fix the measurement problem. It accelerates the decisions built on top of it.

The smarter path isn't more data. It's better questions. What actually causes a repair to fail? Which jobs, if left unresolved, generate a complaint? What's the real reason No Access rates are rising in a particular estate? What are residents reporting vs. what gets logged? These questions require getting beneath the aggregate and looking at individual jobs, individual residents, individual decisions.

This is exactly where Alix is designed to work. Rather than adding another layer of abstraction, Alix brings causal context into the triage process. Before a job is allocated, the team can see what they're actually dealing with: job history, photos, resident vulnerability flags, and communication all in one place.

That changes the quality of decisions at the point they matter most. Repairs managers aren't working from a job number in a spreadsheet. They're working from a real picture of a real situation. The result is better first-time fix rates, fewer repeat visits, and a team that can prioritise with confidence rather than instinct.

The most expensive thing in social housing repairs isn't the operatives or the materials. It's the decisions made with incomplete information. If your current reporting gives you numbers that feel fine but a service that doesn't, it might be worth asking what the data isn't telling you.

If you'd like to see how Alix helps teams get beneath the surface of their repairs data, we'd love to show you.