Are UK firms paying their suppliers faster?
Every large UK company has to publish how quickly it pays its suppliers. We analysed 114,756 of those filings, from 10,238 companies since 2017. The median company's share of invoices paid late has fallen by more than a third — and it is not because companies quietly gave themselves longer to pay.
Key findings
- In 2025, the median large UK company reported that 15% of its invoices were paid outside agreed terms, down from 26% in 2018 (financeb2b analysis of 114,756 UK payment practices reports).
- The median large UK company took 32 days to pay a supplier invoice in 2025, down from 35 days in 2018.
- Among the 4,386 companies that filed in both 2018 and 2025, the median share of invoices paid late fell from 25.5% to 15.6%, and 66% of individual companies improved.
- The median longest standard payment term was 60 days in both 2018 and 2025, so the improvement in punctuality is not explained by lengthening payment terms.
- 11.8% of reports filed for 2025 — 1,404 of them — still showed more than half of invoices paid outside agreed terms.
- 35.2% of companies reporting for 2025 set a longest standard payment term beyond 60 days, and 10.5% beyond 90 days. Long terms are legal and never count as "late".
UK late payment statistics: what nine years of data show
Since April 2017, UK companies meeting two of three thresholds — £54m turnover, £27m balance sheet total, 250 employees — have had to publish a report twice a year on how they pay suppliers. Nine years in, that is a large, consistent, public record of corporate payment behaviour, and it is not a survey: filing is a legal duty and the numbers are the companies' own.
The headline movement is clear. In 2018, the median filer reported that 26% of its invoices were paid outside the agreed terms. By 2025 that had fallen to 15%.
The share of invoices settled within 30 days moved in the same direction, from 55% to 64%, and the share taking longer than 60 days fell from 8% to 5%.
The obvious objection, and why it does not hold
An improvement in "paid on time" can be manufactured without paying anyone faster: lengthen the agreed terms, and the same behaviour becomes compliant. Any honest reading of this dataset has to test that first, because it is exactly what the numbers would look like either way.
It does not hold here. The reports also record each filer's standard payment terms, and across the whole period the median is flat:
| Reporting year | Median longest term | Mean longest term | Share of terms over 60 days | Median share paid late |
|---|---|---|---|---|
| 2018 | 60 days | 69.7 days | 38.4% | 26% |
| 2019 | 60 days | 69.9 days | 37.5% | 23% |
| 2020 | 60 days | 69.8 days | 37.4% | 23% |
| 2021 | 60 days | 69.7 days | 37.6% | 20% |
| 2022 | 60 days | 69 days | 36.4% | 20% |
| 2023 | 60 days | 69.5 days | 36.5% | 19% |
| 2024 | 60 days | 68.5 days | 35.9% | 17% |
| 2025 | 60 days | 67.6 days | 35.2% | 15% |
The median longest standard payment term was 60 days in 2018 and 60 days in 2025. The shortest term is unchanged at 14 days.
A median that sits on a common value can hide a shifting distribution, so it is worth checking the rest of it. The distribution moved slightly the other way: mean longest terms fell from 69.7 to 67.6 days, the share of companies setting terms beyond 60 days fell from 38.4% to 35.2%, and the lower quartile fell from 60 to 52 days. Terms got marginally shorter, not longer.
Companies did not buy their way to better punctuality by moving the goalposts; the absolute time to pay fell too, from 35 to 32 days.
The second objection: is it the same companies?
The number of filers falls over the period, from 7,512 companies in 2018 to 6,548 in 2025. That opens a second way for the trend to be an illusion: if the slowest payers simply dropped out of the dataset — taken over, wound up, or shrunk below the reporting threshold — the average would improve without anyone paying faster.
So we repeated the comparison on a fixed panel: only the 4,386 companies that filed in both 2018 and 2025, matched by company number.
| Group | Companies | Median share paid late |
|---|---|---|
| Filed in both years (fixed panel) | 4,386 | 25.5% → 15.6% |
| Left the dataset after 2018 | 2,546 | 26% in 2018 |
| Joined the dataset by 2025 | 1,762 | 15.5% in 2025 |
The same companies improved. On the fixed panel the median share paid late fell from 25.5% to 15.6%, and the median company improved by 6 percentage points. 66% of individual companies got better, 30.2% got worse. The companies that left were no worse at the start (26% against the panel's 25.5%) and the ones that joined look like everyone else at the end (15.5% against 15.6%). Changing composition explains almost none of the movement.
Why it still does not feel like an improvement
A median is a poor description of a problem that is concentrated in a tail, and this problem is. In 2025, 1,404 reports — 11.8% of all filings that year — still showed more than half of invoices paid outside the agreed terms. Just under a third (32.7%) showed more than a quarter of invoices paid late.
For a supplier, the aggregate does not matter. What matters is which customers they have. If two of your five largest customers sit in that tail, your experience of the last nine years has not improved at all, whatever the median says.
Long terms are legal, common, and not counted as "late"
Nothing in this dataset flags a long payment term as a problem, because contractually it is not one. A company that agrees 90-day terms and pays on day 89 reports as fully compliant. In 2025, of the 8,370 reports stating a longest standard term:
| Longest standard payment term | Reports | Share |
|---|---|---|
| 30 days or fewer | 1,416 | 16.9% |
| 31 to 60 days | 3,998 | 47.8% |
| 61 to 90 days | 2,079 | 24.8% |
| 91 to 120 days | 572 | 6.8% |
| More than 120 days | 305 | 3.6% |
35.3% of reports set a longest standard term beyond 60 days, and 10.5% beyond 90. Punctuality has improved against terms that, for a large minority of filers, were long to begin with.
What this means if you are the supplier
Two practical points follow from the data rather than from opinion.
First, the terms you agree matter more than the punctuality you hope for. The median filer is now reasonably reliable against its terms, so the negotiation that decides when you get paid is the one about the term itself, not the follow-up chasing afterwards.
Second, this record is public and searchable, one company at a time. Before signing with a large customer, you can look up exactly what they filed: their average time to pay, the share of invoices they paid late, and their standard terms. That is a stronger basis for a credit decision than a reference. Search any company on the government's payment practices service.
Where cash timing is the binding constraint, the mechanics of collection are worth getting right: Direct Debit removes the chasing step entirely for recurring amounts, and our payment timing calculator shows when funds actually land once a payment is instructed.
The full series
Every figure behind the charts, including the means the text does not use, so the whole distribution is visible rather than just the measure we chose.
Show the complete year-by-year table
| Year | Reports | Companies | Paid late, median | Paid late, mean | Days to pay, median | Days to pay, mean | Within 30 days, median | Over 60 days, median |
|---|---|---|---|---|---|---|---|---|
| 2018 | 13,419 | 7,512 | 26% | 31.1% | 35 | 37 | 55% | 8% |
| 2019 | 14,999 | 7,690 | 23% | 28.7% | 34 | 36.3 | 56% | 8% |
| 2020 | 13,789 | 7,056 | 23% | 29.1% | 34 | 37.3 | 56% | 8% |
| 2021 | 13,414 | 6,869 | 20% | 26.6% | 33 | 36.3 | 59% | 7% |
| 2022 | 13,068 | 6,664 | 20% | 25.9% | 33 | 35.6 | 60% | 7% |
| 2023 | 13,106 | 6,687 | 19% | 24.9% | 32 | 35.2 | 62% | 6% |
| 2024 | 13,212 | 6,799 | 17% | 23.2% | 32 | 34.5 | 64% | 5% |
| 2025 | 12,721 | 6,548 | 15% | 22% | 32 | 34.5 | 64% | 5% |
Method
Source
Every report published under the UK's payment practices reporting duty, downloaded in full from the Department for Business and Trade's bulk export on 29 August 2026. Published under the Open Government Licence v3.0.
What we did
- Analysed 114,756 reports from 10,238 companies, covering reporting periods ending 2017 onwards.
- Reported medians, not means. The distribution is skewed by a small number of extreme filings, and a mean would overstate the typical company's behaviour in both directions.
- Excluded 49 individual values (not whole rows) whose stated average time to pay exceeded 365 days or was negative, and 5 rows with an impossible reporting date (the raw data contains a small number of evident typing errors, including a period ending in the year 3025).
- Used 2018 as the baseline year, not 2017. The duty began in April 2017, so 2017 contains only a few hundred filings and is not comparable.
- Excluded the current year from all trend statements, because its reports are still arriving.
- The unit of observation is the report, not the company. Filing is half-yearly, so a company filing twice in a year counts twice in that year's medians. The fixed-panel comparison above works at company level instead, averaging a company's filings within each year.
- "Average time to pay" is each company's own average, so our figure is a median of company averages, not a median of individual invoices.
- Only 66% of 2025 reports state a longest standard payment term; the terms analysis covers those reports only. Payment terms of zero days (payable on receipt) are included.
- For context on the baseline: 2017 showed 24% on only 861 filings, so 2018 is both the first comparable year and the highest in the series.
Limitations
- The duty applies only to large companies. Nothing here describes how SMEs pay each other, which is where much late payment is reported anecdotally to occur.
- The figures are self-reported. They are a legal filing rather than an audited account, and definitions of when an invoice is "received" vary between filers — several say so explicitly in their narrative fields.
- The dataset counts invoices, not value. A company can pay thousands of small invoices promptly and be slow on a few large ones without that showing here.
- "Paid late" means outside the agreed terms. It carries no judgement about whether those terms were reasonable, which is why the terms distribution above matters.
Reproducing this
The analysis script is in our repository at scripts/studies/late_payments.py. It downloads the source export, applies the exclusions above and writes the aggregates this page renders, so every figure here traces back to a public file and a published method.
Reuse and citation
You are welcome to cite or reproduce these figures, including the charts, with attribution and a link to this page. The underlying data is Crown copyright under the Open Government Licence and should be attributed to the Department for Business and Trade.
Suggested citation
financeb2b (2026). UK late payment statistics: nine years of official data. Analysis of 114,756 UK payment practices reports. https://www.financeb2b.co.uk/resources/studies/late-payments-uk.php
Download the aggregated figures as JSON — every number on this page, in the form the page reads them, so you can check our arithmetic or chart the series yourself. The full source export is available directly from the government service.
financeb2b research · Analysis by the financeb2b editorial team · Data downloaded 29 August 2026 · Last updated 30 August 2026 · Editorial policy
Getting paid is half the problem
The other half is what happens to the money once it arrives. A Gemba account covers BACS, Faster Payments, CHAPS, SEPA and SWIFT from one balance.
Open a Gemba account