Data quality and TSMs
Somewhere in England this month, a board is being told that overall tenant satisfaction stands at 72 per cent. Heads nod, the figure goes into the annual report and the meeting moves on to development. Nobody asks the awkward question, which isn’t “is 72 per cent good?” but “how confident are we in the plumbing underneath it?”
The Tenant Satisfaction Measures (TSMs) were designed to let tenants and the Housing Regulator compare housing providers on a common footing, comprising 12 perception measures from surveys and ten management information measures calculated straight from your systems. Each half is only as reliable as the data feeding it and, having spent more time under the floorboards of housing data than anyone would wish to, I can report that the plumbing is rarely as sound as the numbers suggest.
When an objective measure isn’t objective
Let’s start with repairs. ‘Completed within target timescale’ sounds like a purely objective measure until you look at where the dates come from. We regularly find jobs closed on the day the operative attends, with the follow-on work raised as a new repair, resetting the clock and invalidating the measure. We find priority codes applied inconsistently between the contact centre and the contractor so the same leak is a seven-day job in one system and a 28-day job in another. The measure is being calculated faithfully but you can’t say the same about the inputs.
Building safety has the opposite problem. The certificates usually exist but they live in a contractor portal, a shared drive or an inbox, or indeed anywhere except the field the TSM is calculated from. One housing provider we worked with was reporting 94 per cent gas compliance from its housing system while the contractor’s records showed 99.6. The homes were safe but the data wasn’t.
Then there are the denominators. Almost every management information measure is a fraction, the bottom of which is your stock list. Garages counted as dwellings, demolished properties lingering on, shared ownership homes misclassified as rentals, all quietly bending every percentage built on top of them. Mrs Miggins’ three phantom garages (see the March 2026 edition of Housing Technology) aren’t just a migration anecdote; they’re now a rounding error in your regulatory return.
Even the perception measures, which look survey-shaped rather than data-shaped, rest on your tenancy records. The sample frame (i.e. the list of tenants the survey company draws from) is taken from your systems so every duplicate tenant, dead phone number and long-departed household skews who actually gets asked. If the only tenants you can reach are the ones who recently updated their details, you’re measuring the satisfaction of your most engaged residents and then calling it the whole population.
“We need to see the working”
The most common trap is the annual panic: leaving the numbers until submission season and then reverse-engineering evidence in April. It works in the sense that a return gets filed but it’s liable to produce figures nobody can defend and a process nobody can repeat. The Housing Regulator has already said that it follows up on ‘outliers’ and engages where it doubts data quality, while an implausibly good score attracts exactly the same attention as a bad one.
The second trap is aiming for perfection. You don’t need 100 per cent clean data, but you do need the fields that feed the 22 measures to be demonstrably trustworthy, with a manageable pile of known exceptions. Chasing perfection across the whole estate is how data quality programmes die.
The third is assuming that somebody else owns the problem. The survey house doesn’t fix your sample frame, the contractor doesn’t reconcile your compliance module and the systems integrator, as ever, unloads the van but doesn’t pack the boxes.
Easily-achievable fixes
The good news is that the fix is methodical rather than heroic. Take each measure and walk it backwards to its source fields then write that lineage down. When the board asks how the number was built, you can say, “Here is the calculation, here are the fields and here are the known exceptions.”
Profile the denominators first. Interrogating the stock list, hunting for zero-bedroom dwellings, duplicated assets and tenures that don’t match the rent charged will improve every measure at once and it’ll be the highest-value cleansing you’ll do all year.
Automate the checks and run them regularly. A simple set of rules, such as completion dates before raised dates, certificates expiring with no renewal booked and complaints open beyond stage timescales, turns an annual archaeological dig into a routine exceptions report so that any drift gets caught in June, not discovered in March.
Use the touchpoints you already have. Every repair appointment and every survey call is a free opportunity to confirm a phone number and tidy the sample frame. It costs nothing and compounds beautifully.
Finally, agree your definitions once, in writing, across the organisation. Half the sector’s problems around complaints data aren’t system problems at all, but three teams holding three divergent opinions about what counts as a complaint.
If you are mid-transformation, fold this work in rather than running it twice. Profiling and cleansing for a migration and profiling and cleansing for the TSMs are the same job wearing different lanyards and doing it once is distinctly cheaper than doing it twice.
Back to page four
As ever with data quality, none of this is glamorous, which is rather the point.
The housing providers who do it will walk into next year’s board meeting with the same slide and a different sentence underneath it: not “satisfaction is 72 per cent” but “satisfaction is 72 per cent and we can show our working”. In today’s new regulatory world, it’s the second half of that sentence that matters.
David Bamford is the commercial director of Migra Data.
For further details, please see: www.migra.co.uk.
