Maladministration rates now travel a long way. They appear in board papers, in trade press headlines, in procurement questionnaires and in the slide a chief executive is handed before an inspection. Very few of the people quoting them have asked how the number was assembled, and none of the places it gets quoted require them to.
That is worth doing, because the answer is more interesting than it sounds.
What is actually published
The Housing Ombudsman publishes more than almost any comparable body in this sector. Every investigated case is written up and put in the open. The decisions archive runs from December 2020, publishes on a fortnightly cycle three months after the final decision date, and names the landlord in every case while keeping the resident anonymous. It can be filtered by category, outcome, tenure, landlord type and the orders made.
Volume has grown quickly with it. The Annual Complaints Review 2024-25 records 7,082 determinations, a 30 per cent increase on the previous year.
None of what follows is a complaint about that transparency. It is the opposite. The material exists, it is public and it is free to read, and that is exactly why it deserves to be handled properly. The issue is that it was published to hold organisations to account, not to be counted, and those are different design goals.
Four things that decide whether a number means anything
The quarterly figures are provisional. The Ombudsman's own Q2 2025-26 landlord report states that all data is provisional and subject to confirmation in the final end-year figures, and directs readers to the Annual Complaints Review for verified numbers. Quarterly figures are the ones that reach a board fastest, and they are the ones most likely to move.
A definition boundary runs through the middle of the series. From April 2025 the categories were renamed. What was reported as property condition became responsive repairs and activity. What was reported as complaint handling became managing relations. A trend line drawn across that boundary is comparing two things that share an axis and not much else. Anyone charting three years of category performance has either handled that transition deliberately or produced a chart that quietly changes meaning halfway along.
The headline rate has no statistics regulator behind it. In the three weeks after the October 2023 annual review, six complaints about the maladministration rate reached the Office for Statistics Regulation. Sector objections, as reported at the time, were that three severity levels are combined into a single figure and that no adjustment is made for organisation size, so a small landlord with a handful of adverse findings can read worse than a large one with many. The OSR declined to investigate, on the basis that the rate is not classed as an official statistic within its remit. That is not a scandal. It does mean the most quoted performance number in social housing carries no external statistical assurance, and anyone republishing it inherits that.
There is no organisation identifier. Landlords are named, and the name is the only handle you get. No registration number, no company number, nothing that survives a rename. In practice the same organisation appears under variants that differ by a word, by a suffix, by a punctuation mark. Others differ enough that they read as entirely separate entities until you look properly. Every count that claims to be per landlord therefore rests on somebody's decisions about which names refer to the same body, and in almost every published analysis those decisions are invisible. They are also the decisions that move the answer most, because merging or splitting one large provider reorders a table.
Why this is a data problem before it is a housing problem
Put those together and one corpus can produce several defensible and mutually contradictory league tables, all of them accurate in the sense that the arithmetic is right.
This is ordinary work if you have done it before. Free-text sources with no stable identifiers, definitions that shift under you mid-series, provisional figures that firm up later, and a headline metric that mixes severities: none of that is unusual in national data. It is what national data looks like before someone does the unglamorous part. Our background is in delivering healthcare data services at national scale, where the same problems arrive with statutory consequences attached, and the discipline that survives from it is simple. Assume the source is inconsistent until proven otherwise, and never let an internal check stand in for the document itself.
Two examples of what that changes in practice, since the sector's own objections named them. Severity levels are kept apart rather than summed, so a severe finding never disappears into an average. And small-sample rates are shrunk toward the sector mean using empirical Bayes, which is the standard treatment for exactly the complaint raised in 2023: eight adverse findings out of ten is not the same evidence as eight hundred out of a thousand, and a raw percentage treats them identically. A score built that way reports its own confidence alongside the number.
Three questions to ask anyone selling you sector data
Not as a challenge. As the same due diligence you would apply to any other management information.
Which version of the figures is this, provisional or verified, and as at what date? How is the category change of April 2025 handled in anything spanning it? And how were landlord names resolved to organisations, given the source carries no identifier?
A supplier who has done the work will answer all three quickly, because they had to solve them to produce anything at all. A supplier who has not will answer the first and change the subject. The questions cost nothing and they are the difference between a number you can put in front of a board and a number you can put in front of a board once.
The sector pattern in the published record is free for anyone to read, and worth reading. Where a particular organisation sits inside it, once the naming, the versioning and the severity split are handled properly, is a different conversation and not one for a public page. If that is useful, book a private readout. Twenty minutes, and no deck.
Figures and publication practice described here are drawn from Housing Ombudsman publications and contemporary trade reporting, current at 4 August 2026. Nothing here names an individual landlord or characterises any organisation's performance.