Forecast late payers from the pattern of their lateness, not from an average. In Adfin's payment data, about half of late payments arrive within a week of the due date, while a small tail runs for months and pulls the average up to around four weeks. Use your typical delay for the base case, a slower figure for the downside, and take very old debt out of the forecast altogether.
In this article
The short version
- An average days-late figure puts money in the wrong weeks, because a few very late invoices pull it up.
- In Adfin's data the typical late payment arrived 8 days late, and about half within a week.
- Forecast each open invoice at your median delay for the base case and a slower figure for the downside.
- Put direct debit receipts in on their collection dates, and take anything over 90 days out of the forecast.
- Stress test by changing one assumption, running a bad month and working back from failure.
Why an average delay gives you the wrong picture
The usual shortcut is to add your clients' average delay to every due date. It's quick, but it tends to put money in the wrong weeks, because lateness isn't spread evenly.
Adfin's own data on card, pay by bank and bank transfer payments that arrived after their due date shows the shape clearly:
- The typical (median) late payment arrived 8 days after the due date, but the average was about 28 days.
- 48% of late payments arrived within a week of the due date, and 80% within a month.
- 7% ran past 90 days, and about 1% took ten months or more.
The figures come from a mix of mostly small UK businesses over about two years, so it's best to treat them as a guide to the shape rather than a benchmark for your own clients.
If you put 28 days on every late invoice, your first fortnight will look emptier than it turns out to be. If you use 8 days on its own, you've left out the invoices most likely to hurt you. Both numbers belong in the forecast, doing different jobs: the typical delay sets your base case, and the tail tells you how much is at risk.
Direct debit doesn't belong in this pattern at all. You choose the collection date and the money follows on a set timetable, so those receipts go in on the day you expect them. If you blend them into a days-late average, you're mostly measuring bank processing time, not your clients.
How to build the receipts line
The receipts line is the only part of your forecast where someone else decides the timing, so it's worth building it invoice by invoice.
- List every open invoice with its due date and amount. Your aged debtor report is the starting point.
- Split them by how each client pays. Direct debit invoices go in on their collection dates. Everything the client has to act on goes into your lateness model.
- Work out your own pattern. Take your last 100 to 200 paid invoices and calculate days late for each (payment date minus due date). Note the median, and the share that ran past 30 and 90 days. It doesn't need to be exact. For your biggest clients, use their own history, because one client's habit can outweigh your whole average.
- Set a base case. Put each open invoice in at its due date plus your median delay, or that client's usual delay if you know it.
- Set a downside. Use a slower figure, such as the delay that a quarter of your late payments exceed. In the figures above, that's 23 days.
- Take anything over 90 days out of the forecast. Hold it on a separate recovery list with its age, the date you last made contact and your next step, and count it only in your upside.
That last step is the one people tend to resist. An invoice that's 200 days old still has a real chance of being paid, but it hasn't got a predictable date, so leaving it in as a receipt in week three makes every week after it look better than it is.
How to test the forecast
Thirteen weeks, rolled forward each week, is a popular length for a working forecast: long enough to show a quarter's payroll and VAT, short enough to keep accurate. Formal guidance only sets a horizon for going concern assessments, which the Financial Reporting Council expects to cover at least a year.
The same FRC guidance warns that when you're preparing a forecast, you can be biased by your own assumptions. On the receipts side that usually means optimism about when a particular client will pay, and it tends to attach to the clients you like most. The guidance describes three ways to test for it, and each one works well on late payers:
- Change one assumption at a time. Push your median delay out by a week and see which weeks go negative.
- Run a bad scenario. Try your three largest clients all paying 30 days late in the same month.
- Work backwards from failure. Find the pattern of late payment that would leave you unable to pay staff, then check how far your current book is from it.
How to keep it up to date
A receipts forecast goes stale quickly, because it depends on knowing which invoices have been paid. If payments aren't matched to invoices until days after they arrive, or reminders only go out when someone remembers, the forecast drifts away from your bank balance, and it's easy not to notice.
Most of that upkeep is routine, which makes it easy to hand to software. Adfin matches each payment to its invoice as it arrives, including part payments and missing references, and syncs it back to Xero or QuickBooks, so your list of open invoices stays current. Its AI Customer Agents follow up each overdue invoice by email, text or WhatsApp, within rules you approve. And moving regular clients onto direct debit takes their invoices out of the guesswork, because you know the collection date in advance.
Common questions
Can I use an average days-late figure in a cash flow forecast? On its own it's likely to mislead you, because a few very late invoices pull the average up. Use your median delay for the base case and a slower figure for the downside.
What's a realistic assumption for how late a client will pay? Work it out from your own last 100 to 200 paid invoices. In Adfin's data the typical late payment arrived 8 days after the due date.
Should direct debit payments be forecast the same way? No. You set the collection date, so direct debit receipts go in on that date rather than into a lateness average.
How do I work out days late for my own invoices? For each of your last 100 to 200 paid invoices, subtract the due date from the payment date. Note the median and the share that ran past 30 and 90 days.
What do I do with invoices more than 90 days overdue? Take them out of the forecast and track them on a separate recovery list. Count them only in your upside case.
How do I stress test a cash flow forecast? Change one assumption at a time, run a bad scenario such as your largest clients paying late together, and work backwards from the pattern that would leave you unable to pay your costs.
Sources
- Financial Reporting Council - Guidance on the Going Concern Basis of Accounting and Related Reporting (accurate as of October 2026) https://media.frc.org.uk/documents/Guidance_on_the_Going_Concern_Basis_of_Accounting_and_Related_Reporting.pdf
- Bacs - Direct Debit (accurate as of October 2026) https://www.bacs.co.uk/bacs-schemes/direct-debit/
- Adfin - Pricing (accurate as of October 2026) https://www.adfin.com/pricing
- Adfin - Credit control (accurate as of October 2026) https://www.adfin.com/credit-control
