Methodology

Where the data comes from

English councils are required to publish their spending over £500 to individual suppliers. We ingest each council's own published disclosure files directly — we don't estimate or infer transactions.

How we clean and normalise it

Supplier names are merged when they differ only by a trailing company suffix, punctuation or capitalisation — so "ACME LTD", "Acme Limited" and "The Acme Company" are counted as one supplier rather than three. We deliberately do not merge names that differ by more than that: doing so risks attributing one company's spending to another, which would be a worse error than listing it twice. Each transaction is also assigned to one of 15 standard service categories using keyword matching against its description, so spending can be compared across councils even though each council describes its own transactions differently.

How we count repeated payments

A council will often pay the same supplier the same amount on the same day more than once — separate invoice lines for separate pieces of work. Until July 2026 our records collapsed those into a single row, because we treated supplier, date, amount and description as enough to identify a payment uniquely. It isn't. Retaining them separately added 25.8 million payment lines, taking the dataset from 51.6 to 77.7 million.

We want to be exact about how much of that we can prove. Of the 26.2 million repeated rows involved:

• 25% carry distinct references. Where a council publishes its own invoice, payment or line numbers, those numbers show the rows are separate payments.
• 15% carry an identical reference across the repeats.
• 60% come from councils that publish no per-payment identifier at all. 89 of the 131 councils publish none; for Newham, Lambeth, Essex, Hampshire and Gateshead it is all of their data.

The clearest test is Croydon, which does publish line numbers. On one day it paid Wates £113.95 for "Responsive Repairs - Building" 3,678 times, and the line numbers confirm 3,678 separate repairs — £419,108 of work that our old records stored as a single £113.95 payment.

So where the premise can be tested, it holds. Where a council publishes no identifier we apply it anyway, which is an inference rather than a proof — those totals are our best reading of the disclosure, not a figure the council itself can confirm. We would rather state that plainly than present the whole 50% increase as settled fact.

Coverage and update cadence

We currently track 131 English councils, covering 78.7 million published payment lines. Councils publish disclosure files on their own schedule and in their own formats, so how recent the data is varies council by council — some are updated within the last few months, others less frequently, depending on what each council has published. Each council's dashboard shows the date range its data actually covers, so you can see exactly how current it is rather than assume a fixed cadence.

Known limitations

These are the specific, known ways our figures can differ from a council's actual spending. We publish them because you should be able to judge the numbers for yourself.

• For most councils, repeated payments can't be independently verified. As set out above, we retain same-supplier, same-day, same-amount payments as separate lines. For the 89 councils that publish no per-payment identifier there is nothing in the source file that confirms whether such rows are genuinely separate payments or the same one listed twice. Councils that itemise many small identical payments — care placements, foster carer allowances, repeat hire — are where this matters most.
• Some councils publish one payment as many apportioned lines. Where a council splits an invoice across cost centres, its file may repeat the full invoice value on every line rather than the split amount. Reading the wrong column then counts that invoice once per line. We found this in Cornwall's data on 1 August 2026 and corrected it, which took Cornwall's published total from £28.7bn to £7.0bn; we are checking the remaining councils whose files show the same shape. Where a council provides a per-line amount we now use it.
• Publication thresholds differ between councils. The statutory duty covers spending over £500, but many councils publish below it — some at £250, some every transaction including purchase-card spend. For over a dozen councils in this dataset more than half of all published transactions fall under £500. A council that publishes more fully will therefore appear to have more transactions than one that publishes only the statutory minimum, and direct comparisons of transaction counts between councils are not like-for-like.
• Duplicate suppliers still exist. Suffix variants are merged, but source files also truncate long names at different widths and contain spelling errors, so one supplier can still appear under several near-identical entries.
• Source files differ in structure between councils, which occasionally leads to gaps or formatting quirks in fields like description or date.
• Transaction records carry limited context beyond supplier, amount, date, and category — they show what was paid, not the full contract or decision behind it.
• Figures reflect what each council has chosen to disclose, not an independently audited account.