Across 59,777 invoices your customer pays themselves, by card, wallet, open banking or bank transfer, 65.7% arrive on or before the due date, and when one is late the median delay is 8 days (Adfin platform data). Those figures describe businesses collecting through Adfin, mostly small practices billing small recurring fees, so read them as a benchmark for your own book and not as a national statistic.
In this article
The short version
- 22.0% of customer-initiated payments land on the due date itself, on 59,777 paid invoices (Adfin platform data).
- Among the 20,533 late ones the median delay is 8 days and the mean 28.4, so an average alone hides a long tail of old debt (Adfin platform data).
- Of 1,359 repeat payers, 27.4% were never late once, and the worst 24.3% account for 51.9% of all late payments (Adfin platform data).
- Direct debit collects 97.0% first time and 98.7% eventually, on 269,518 collections attempted July 2024 to mid-July 2026 (Adfin platform data).
- A due date you set on a Saturday or Sunday is paid on time about six percent less often than a weekday one (Adfin platform data).
- Longer terms go with better on-time payment, 59.5% on same-day terms against 76.4% at 8 to 14 days, and 5.5% of invoices are issued already past their due date (Adfin platform data).
What this report measures, and whose payments it describes
Adfin can see, one invoice at a time, when a customer actually paid: the timestamp, the method, the due date, the retry, the mandate. Almost nobody in the UK holds that join, and the population it covers is Adfin's own book, weighted heavily towards small accountancy practices billing small recurring fees, with half of all customer-initiated invoices under £180 (Adfin platform data). So use each figure as a comparison for your own book and not as a national one. If your work goes out as £40,000 engineering jobs on 60-day terms, the shape of a finding may hold for you and the level almost certainly won't.
The window is invoices created between 16 June 2024 and 17 August 2026: 482,913 paid payment requests raised by more than a thousand UK businesses, all in sterling (Adfin platform data). Every table below carries the invoices behind it, so you can see what each figure is a share of. They split into three groups that can't be added together:
| Population | How it was paid | Share of paid invoices | What its clock measures |
|---|---|---|---|
| Direct debit | Bacs collection under a mandate | 62% | The Bacs settlement cycle |
| Imported | Paid to the business directly, recorded through an accounting connector | 26% | The business's own reconciliation lag |
| Customer-initiated | Card, Apple Pay, Google Pay, open banking, bank transfer | 12% | Your customer deciding to act |
Adfin platform data.
A direct debit enters Bacs a median of two days before the due date and settles five days later, so it lands three days after the due date when nothing has gone wrong, and only 0.20% of collections read as on time. Blend that with the rest and you have published the Bacs calendar as customer lateness, so nothing here mixes the three. Every on-time and days-late figure below is customer-initiated only, the closest thing to your own customers deciding when to act. The workings are in how we measure UK payment behaviour.
Your customers pay on the due date, not before it
The date you put on the invoice is doing more work than anything else here. Of the 39,244 on-time customer-initiated payments, 33.5% arrived on the due date itself and 40.5% in the final two days, and 22.0% of all 59,777 payments landed exactly on the day they were due (Adfin platform data). The median on-time payment came 3 days early.
The other half of your book pays on receipt. 23.0% of customer-initiated payments arrive within an hour of the request, a quarter within two hours, 42.7% within a day and 67.5% within a week, with a median of 53 hours (Adfin platform data). So two habits share one due date, and nothing here can tell you a reminder produced the second one, because Adfin's warehouse holds no record of chasing events. Payments cluster tightly on the date you set, and that date is the most useful lever you have on this page.
Arrival looks like office hours: 90.9% of customer-initiated payments land Monday to Friday and 64.2% between 09:00 and 16:59 UTC, though about one in five arrives at or after 17:00 UTC or before 06:00 (Adfin platform data). So a fifth of your cash moves outside anybody's working day.
When a payment is late, how late is it?
34.4% of those 59,777 payments arrived after the due date, so if your own late rate is near a third you're in ordinary company. Here is how late the 20,533 were.
| Days late | Value |
|---|---|
| Mean | 28.4 |
| Median | 8 |
| 25th percentile | 3 |
| 75th percentile | 23 |
| 90th percentile | 68 |
| 99th percentile | 321 |
Adfin platform data, 20,533 late customer-initiated payments.
The mean is three and a half times the median, so an average on its own would misdescribe almost every invoice here. Half of your late payments behave like admin: 14.1% are late by one day, 27.3% by three days or fewer, 48.3% by a week or less (Adfin platform data). The mean is carried by the other end, where 7.1% run past 90 days and the 99th percentile is 321 days.
If you're deciding where your chasing time goes, that split is where to start: a week spent on the one-day-late group buys you very little, and your money is at risk in the 7.1%.
If you want an external comparator, Xero puts UK small businesses at 7.3 days late on average. The nearest figure here is your median of 8 days, and the two aren't directly comparable: different businesses, different definition, different denominator.
Late payment belongs to a minority of your payers
Lateness isn't spread evenly across your customers. Take every business-customer pair with at least five customer-initiated payments, 1,359 of them covering 12,863 payments (Adfin platform data).
| Repeat payers | Share of pairs |
|---|---|
| Never late once | 27.4% |
| Late on every single payment | 10.3% |
| Late 80% of the time or more | 24.3% |
Adfin platform data, 1,359 business-customer pairs with five or more customer-initiated payments.
That worst quarter generates 51.9% of all late payments in the group (Adfin platform data). So on a two hundred client ledger, roughly fifty names produce half your chasing work and about fifty-five pay on time without you doing anything. You can run the same cut on your own history: rank every client with five or more payments by how often they missed.
One limit, and it matters. A pair only enters this cut after five customer-initiated payments, and a customer who paid reliably by card five times running is exactly the one most billers would have moved onto a mandate. So the sample leans towards customers who resisted that. The concentration is real in this group, and the exact proportions won't transfer to your whole book.
The payment method league table, and its caveat
Here's the table you came for, with its caveat attached.
| Method | Paid on or before the due date | 95% confidence interval | Invoices measured |
|---|---|---|---|
| Apple Pay | 72.86% | 71.93 to 73.79 | 8,751 |
| Google Pay | 70.00% | 67.93 to 72.07 | 1,880 |
| Card | 65.13% | 64.65 to 65.61 | 38,284 |
| Open banking | 62.00% | 60.94 to 63.06 | 8,010 |
| Bank transfer | 57.85% | 56.04 to 59.66 | 2,852 |
Adfin platform data, 59,777 paid customer-initiated invoices across 1,042 businesses. Card, open banking and bank transfer have been restated upward from 64.2%, 60.4% and 55.2% published earlier; Apple Pay and Google Pay are unchanged.
The fifteen percent spread is mostly about which businesses turn a method on and which of their customers reach for it, and not about the rails. Hold the business constant, compare methods inside one book, and the gaps collapse to under 1.5 percent with the split between businesses close to a coin flip (Adfin platform data). So adding Apple Pay to your checkout is reasonable because some of your customers prefer it, and this data can't promise you it changes when they pay. The within-business test is in the method page.
Once a payment is late, your customer's method stops separating anything: the median delay is 8 days for card, wallets and open banking alike, and 10 days for bank transfer (Adfin platform data). Wallets do arrive earlier when they arrive on time, the typical Apple Pay payment two days before the due date and Google Pay one day before, which says something about when people choose to act and nothing about clearing speed.
That pattern repeats everywhere here. Invoice size, payment method and sector each move the probability of missing a due date by a few percent, and not one of them changes the length of the delay: your median late payment is 8 days late however you cut it, including across every invoice size band, as article 093 works through.
Direct debit: what fails, and what a retry is worth
Your direct debits are measured against Bacs and never against the due date. On 269,518 collections first attempted between July 2024 and mid-July 2026, 96.97% succeeded first time and 98.67% were collected eventually, while 2.58% saw a failed attempt and 60.13% of those were still collected in the end (Adfin platform data).
That cohort matters, because roughly 11,800 Bacs attempts are in flight at any moment and an in-flight collection isn't a failed one. Without a maturity gate the same query reads 93.74% and 95.31%, a measurement of the settlement calendar instead of your reliability.
When a collection does fail, the value of trying again decays quickly.
| Attempt | Succeeded | Attempts measured |
|---|---|---|
| First | 96.97% | 269,518 |
| Second | 46.92% | 6,371 |
| Third | 24.96% | 2,921 |
| Fourth | 19.14% | 1,886 |
| Fifth | 16.39% | 775 |
Adfin platform data, 281,471 direct debit attempts on 269,518 invoices, all attempted before 15 July 2026.
So if you're writing a retry policy, take the second attempt and think again before the fourth, when your invoice probably needs a conversation instead. The gap between attempts is a consistent 6 days, the Bacs re-presentation cycle, so there's no timing choice in there for you to optimise, and none of this is chasing: these are automated re-presentations against a mandate. One caveat before you read the failure rate as a trend: it has crept up from about 1.3% to 2.8% as the book has grown, while first-attempt success stayed flat.
What a mandate predicts about an invoice
The strongest single predictor of whether your invoice gets resolved is whether a direct debit mandate already existed for that customer when you raised it. Measured on 152,689 invoices created in the seven months to 31 January 2026, so each has had at least six months to resolve:
| Mandate in place when the invoice was raised | Paid | Still open six or more months later |
|---|---|---|
| Yes | 96.98% | 1.31% |
| No | 90.25% | 5.36% |
| Median invoice value | £126 with a mandate | £180 without |
Adfin platform data, 98,049 invoices with a mandate and 54,640 without. Mandate coverage across that cohort was 64.2%.
Four times as much of your money stays stuck without one. This cut uses resolution status and not a days-late clock, so the Bacs problem doesn't touch it.
It isn't a controlled experiment either, and the difference in median value is the clue: mandated invoices are smaller, and they go to customers who already agreed to a mandate, so some of that reliability was there before the paperwork.
Coverage is still a number you can pull today: what share of the invoices you raised last month had a mandate behind the customer?
The due dates you set yourself
Two findings here are about your own invoicing and not about your customers, so you can act on both this week. A weekend due date costs you about six percent of on-time payment. Saturday runs at 60.38% and Sunday at 60.59% against 65.80% to 67.15% Monday to Friday, and the weekend confidence intervals don't overlap any weekday interval (Adfin platform data, 59,777 invoices, 4,372 due on a Saturday and 3,611 on a Sunday). The mechanism is probably mechanical: a Saturday deadline means Friday was your customer's last chance to act at work. Nobody has tested whether moving the date causes the improvement, so treat it as a clean association and go and look at where your own due dates land.
Then terms. You set those too, and the result runs against what you might expect.
| Terms set | Paid on time | Median days late, when late | Invoices measured |
|---|---|---|---|
| Due before the invoice was issued | 0.00% | 42 | 3,271 |
| Due same day | 59.49% | 5 | 14,800 |
| 1 to 7 days | 70.91% | 7 | 23,493 |
| 8 to 14 days | 76.40% | 9 | 8,554 |
| 15 to 21 days | 73.47% | 10 | 3,159 |
| 22 to 30 days | 75.75% | 9 | 5,678 |
| 31 to 60 days | 76.30% | 11 | 616 |
Adfin platform data, 59,777 paid customer-initiated invoices. Terms are inferred from the gap between the invoice date and the due date.
Giving your customer more time doesn't appear to make them later. On-time payment rises from 59.5% on same-day terms to 76.4% at 8 to 14 days, then flattens around 74% to 76% out to 60 days. Read that as selection and not as advice: businesses that set proper terms tend to run a tighter ledger anyway, and the trade-off is in the same table, since longer terms bring a longer delay on the ones that do go late.
The first row is the self-inflicted one. 5.5% of customer-initiated invoices, 3,271 of them, went out with a due date already in the past, so not one was ever on time and their median delay is 42 days. Much of that will be catch-up or back-dated invoicing, and the warehouse can't tell those apart from a live invoice raised already overdue. If your own ledger carries a share like that, it's the cheapest thing on this page for you to fix.
December is real, August is not
A raw monthly series is confounded by Adfin's own growth, so both tests below hold the same businesses across three consecutive months. That's the version you can trust.
On a fixed panel of 70 businesses, on-time payment ran at 66.07% for November 2025 due dates, fell to 61.90% in December and recovered to 66.01% in January, across 7,071 payments (Adfin platform data). A 4.2 percent dip on a fixed panel can't be growth or mix, and at about 2.8 standard errors chance is unlikely. That's one December measured cleanly and not evidence about every December, so if you're planning December cash, plan for a few points of slippage and a January recovery.
The same test across the summer finds nothing: on a fixed panel of 25 businesses, July 2025 ran at 74.37%, August at 72.96% and September at 77.75%, with the August gap well inside noise (Adfin platform data, 2,788 payments). Everybody assumes there's a summer slowdown in getting paid, and in this book there isn't a measurable one, so if you've been blaming August for your cash position, the cause is likely somewhere else.
The relationship-age paradox
The most surprising thing in this data comes with a warning attached instead of a lesson. Follow one of your customer relationships from its first invoice onwards and direct debit's share of paid invoices roughly doubles: 36.29% in month one, 67.11% by month two, 74.36% by months seven to twelve, while the customer-initiated share falls from 25.82% to 6.81% (Adfin platform data, 482,910 paid invoices). Mandate adoption happens over the first six months of knowing someone.
Over the same months, the on-time rate inside the customer-initiated channel falls: 67.57% in relationship month one, 63.66% in month two, 61.70% across months four to six, recovering a little to 65.42% by months 13 to 24 (Adfin platform data, 59,773 paid invoices).
So the channel looks like it's deteriorating while the relationship improves. Both figures are correct, and they describe different populations: as a relationship matures your reliable payers move onto a mandate and leave the customer-initiated channel, so what remains is a selected residue of customers who won't or can't be put on one.
Nothing here supports "customers get worse over time". At business level the opposite is measured: your overdue rate roughly halves over your first seven months of collections, from 23.01% to 10.15%, on a balanced panel of 292 businesses and 43,081 paid invoices (Adfin platform data). That counts whether an invoice went overdue at all and not how many days late it was, a binary flag being the only measure here that can safely span all payment methods.
Two warnings come with the pair, both about your own reporting. Any channel-level rate you track can move opposite to the underlying outcome when the channel's population shifts underneath it, so a card on-time rate on your dashboard can drift down while your collections improve, purely because your best payers moved onto mandates. And every customer-initiated figure here is a segment defined after the fact by how the invoice ended up being paid, so its composition isn't stable over the life of a relationship.
Who you bill for, and how much you bill
Two segments in Adfin's book are spread across enough separate businesses to publish, and one is probably yours.
| Who raised the invoice | Paid on time | Median days late, when late | Invoices measured |
|---|---|---|---|
| Small accountancy practices | 59.87% | 10 | 27,483 |
| Other small businesses | 72.32% | 5 | 14,784 |
Adfin platform data. 661 businesses in the first row with no single one above 5.8% of the cell, 325 in the second with the largest at 13.0%.
A twelve percent gap, and when the accountants' clients are late they keep you waiting twice as long. If you run a practice, plan around the lower figure; if you bill other small businesses, the higher one. Professional fees being lower-salience and often billed monthly is a plausible explanation, untested here. Adfin's other segments each lean on one or two large customers, so a rate for them would describe those customers and not a sector.
Invoice value moves the probability of lateness by about ten percent from the smallest band to the largest, and explains 0.23% of the variance in whether a payment misses its due date, falling to 0.011% inside one book. What value predicts strongly is your mandate coverage: it halves from 74.1% under £100 to 37.4% at £2,500 and over (Adfin platform data), which has its own page in why your biggest invoices are the ones still collected manually.
What this report can't tell you
Six limits, stated because a report that hides them wouldn't deserve your trust.
- One platform's book, and not a national sample: weighted towards small accountancy practices billing small recurring fees, median customer-initiated invoice £180, only 1% above £3,250.
- Every lateness figure is conditioned on the invoice having been paid, because lateness needs a payment date. The invoices nobody ever paid are missing, they skew larger, and so every rate here understates true lateness.
- Most differences between groups are composition. Inside one business's own book the method gaps shrink to under 1.5 percent and three quarters of the invoice-size effect disappears.
- There's no chasing data at all, so nothing here knows what you sent or when. Which channel and which send hour work are questions this data can't answer, and the clustering on the due date can't be credited to a reminder instead of your customer's own diary.
- Payment terms are inferred from the gap between invoice date and due date, since nothing records what was negotiated.
- The two most recent months are dropped from every trend, because recent invoices still unpaid are structurally absent and make the newest months look artificially good.
The exclusion gate, the definitions and the disclosure rules are all in how we measure UK payment behaviour. If you want to argue with a number above, start there.
Adfin intends to restate this report annually and to date each edition to its snapshot instead of presenting the figures as fixed. This one was computed from a warehouse snapshot taken at 07:04 UTC on 18 August 2026.
Common questions
How quickly do UK businesses get paid? In Adfin's book, 65.7% of invoices that a customer paid themselves arrived on or before the due date, and the median payment came 53 hours after the request went out. Among the late ones the median delay was 8 days. That population is mostly small UK businesses and accountancy practices billing small recurring fees, so treat it as a description of them rather than a national statistic.
What percentage of invoices are paid late? 34.4% of the 59,777 customer-initiated invoices measured here were paid after the due date. Direct debit invoices are counted separately, because a Bacs collection settles a median of three days after the due date by design and would make any blended figure meaningless.
How many days late is a typical late invoice? The median is 8 days and the mean is 28.4, measured on 20,533 late payments. The mean is pulled out by a tail where 7.1% of late payments run past 90 days, so the median is the fairer single number. 14.1% are late by exactly one day.
Which payment method gets paid on time most often? Apple Pay at 72.86%, then Google Pay at 70.00%, card at 65.13%, open banking at 62.00% and bank transfer at 57.85%. Most of that spread is about which businesses offer which method and which customers use it: compare methods inside one business's own book and the gaps fall below 1.5 percent.
Does a weekend due date get paid later? Yes, by about six percent. Invoices due on a Saturday are paid on time 60.38% of the time and Sunday 60.59%, against 65.80% to 67.15% Monday to Friday. The confidence intervals don't overlap, though moving a due date hasn't been tested as a cause.
Is late payment worse in December? It was in December 2025. On a fixed panel of 70 businesses, on-time payment fell from 66.07% in November to 61.90% in December and recovered to 66.01% in January. One December measured cleanly isn't evidence about every December, and the same test found no measurable August effect.
Sources
- Xero Small Business Insights — UK data, April to June 2024 (accurate as of August 2026)
This article describes patterns in Adfin's own platform data and is not legal or financial advice. The figures are observational: they describe one platform's book of mostly small-ticket UK invoices, and no figure here establishes cause. Last updated August 2026.
