"I can't breathe properly in the bedroom." Is anyone reading that?

That sentence appeared in a tenant survey. Not a formal complaint. Not a repair request. A survey comment, filed and forgotten in a spreadsheet somewhere.

Under Awaab's Law, it's a compliance trigger.

The problem most housing associations have isn't ignorance of safety risk. It's lag. Safety issues surface once they've become complaints, press stories, or Ombudsman cases. By then, the clock has long since started — you just didn't know it.

When the clock starts isn't when you think

Awaab's Law doesn't start the 24-hour emergency clock or the 10-working-day damp and mould window when a tenant submits a formal repair request. It starts the moment they raise a hazard on any channel.

That includes TSM surveys. Transactional surveys. Emails. Call notes. Live chat. Social media. And Phase 2, due in 2026, extends those requirements beyond damp and mould to fire, electrical, and structural risks.

Which means every free-text field in every feedback system you run is now legally material. And if you're still monitoring that manually, you're not being diligent. You're being dangerous.

The data problem hiding in plain sight

In our TSM benchmark data, a single mention of security knocks satisfaction down by around 36 points. Damp and mould shows up in roughly 4.9% of comments and drags scores into the mid-30s when it appears. Often paired with health language: asthma, breathing difficulties, kids.

That crossover — damp or mould plus health mentions — is where regulatory, legal, and reputational risk concentrates. And it almost never arrives via a formal complaint first.

Tenants don't write "I wish to formally report a Category 1 HHSRS hazard." They write "black spots in the bathroom" and "can't breathe properly in the bedroom." The hazard is there. The word "mould" sometimes isn't.

This is why a keyword search isn't enough. You need a model that understands what tenants are describing, not just what words they use.

What an early warning system actually needs

The core of a tenant safety early warning system is straightforward in principle: scan free-text feedback across every channel, detect hazard language the moment it appears, and route it for action inside regulatory timeframes. Three building blocks make that work.

Unified data capture. Every feedback source, whether it lives in one platform or not, needs to flow into a common store with tenant ID, date, channel, and free-text preserved. If a hazard mention arrives via email and the relevant property history is in your housing management system, those two things need to talk to each other.

A single hazard dictionary. A defined list of phrases, symptoms, and implied risks mapped to HHSRS hazards and your internal categories. Not just "mould" but "walls always wet," "kids' asthma worse," "door never locks." Pre-trained against UK housing language, not generic support ticket vocabulary.

Audit-ready metadata. Every potential hazard mention needs a durable record: when it was raised, how it was classified, when it was reviewed, what action followed. Without that lineage, you can't prove compliance to the Regulator, the Ombudsman, or in court — even if your teams did exactly the right thing.

Low volume is not low risk

This is the assumption that gets housing associations into trouble.

Safety doesn't behave like repairs or communication. Most satisfied tenants never mention it at all. Those who do skew heavily toward the very dissatisfied end of the spectrum. In our benchmark data, tenants who mention safety are nearly four times more likely to be very dissatisfied than very satisfied.

That 4:1 ratio changes how you should read your own data. A handful of comments about security doors or mould is not comfort. It's a signal that you may only be hearing from the most distressed residents.

Benchmarking helps you calibrate this. Comparing your rate of damp and mould mentions against sector norms tells you whether "we get some complaints" means "we're broadly typical" or "we're running three times the sector average in pre-1970 low-rise blocks." That's the difference between a maintenance issue and a board-level risk.

Getting from spreadsheets to something robust

This doesn't have to be a five-year transformation. A practical roadmap gets you there in roughly 12 months if you keep each phase tightly scoped.

Start with a 90-day pilot on historic data. Take one or two years of TSM and complaint comments, run them through a housing-specific analytics platform, and validate the hazard classifications against known cases. You'll get a realistic false-negative rate and almost certainly uncover patterns you've missed — mould clustering around certain building types, specific contractors, particular postcodes.

The following 90 to 120 days: integrate live channels, stand up basic workflows, build a safety dashboard in Power BI. Daily hazard digests, emergency alerts, automatic reminders as statutory deadlines approach. Focus on damp, mould, and health first.

The remaining months are about deepening coverage as Phase 2 comes into force, refining models based on case outcomes, and getting the dashboards into board packs and risk registers.

By the end of the year, you're not hoping someone spots "I can't breathe properly in the bedroom" in a spreadsheet. You're running a live, auditable system that treats every tenant comment as a potential compliance signal — and can prove it.

The value of feedback is inversely proportional to its latency. In safety terms, that's not just an insight principle. It's a legal one.