Research · Collections

Does automated follow-up get overdue invoices paid? The evidence

What independent data shows about late payment in the UK and US, what automated invoice follow-up changes, what it does not, and how to measure it.

, Founder and Managing Partner9 min read

Yes, for most overdue invoices. The strongest independent evidence, a randomised trial with business debtors, found that a simple reminder raised the probability of payment by about 25 percentage points, and earlier reminders produced faster payment. Follow-up does not fix disputes, missing approvals or insolvent customers. Older debt needs triage, not more emails. What follows sets out the data, ranked by how far it can be trusted.

We draw on three kinds of evidence and label each. Government and credit insurer data comes first. Data from accounting platforms on their own users comes second. Vendor case studies come last, marked as vendor-reported, because they are selected successes.

How big is the late payment problem in the UK and US?

Late payment is a large, persistent cost in both markets. In the UK it is estimated to cost the economy almost £11bn a year. In the US, about seven in ten companies report late payment from business customers.

The UK figures come from the Department for Business and Trade (2025). Over 1.5 million businesses, 28% of the total, are affected each year. About £26bn is owed in late payments at any time, an average of £17,000 per affected business. The department links late payment to 14,000 business closures a year. Firms that spend staff time chasing payment report an average of 86 hours a year doing so.

The credit insurer Atradius gives a view from inside B2B trade. Its 2026 UK barometer, a survey of 210 UK firms, found that around two thirds report delayed payments. Late payment affects about 25% of invoiced B2B turnover. The report also notes that UK debts which fall behind are more likely than elsewhere in Western Europe to drift into long-term arrears.

For the US, the Atradius 2026 US report states that late payment affects an average of 22% of B2B receivables, and that US firms often lose 1% to 2% of B2B invoices to bad debt. Smaller firms feel it most. The QuickBooks 2025 Small Business Late Payments Report, a survey of 2,487 US firms with up to 100 employees, found that 56% were owed money on unpaid invoices, $17,500 on average, and 47% had invoices more than 30 days overdue. This is platform research, but it is the largest recent US sample.

Policy is moving too. The UK government consulted in 2025 on measures that include payment terms beyond 60 days, published its response in March 2026 and intends to legislate. Finance leaders should expect closer scrutiny of payment practice on both sides of the ledger.

Why do invoices go overdue?

Most invoices go overdue for one of two reasons. Either the customer is short of cash, or something administrative stops the payment. The second group is where follow-up does most of its work.

In the Atradius UK survey, the top reasons given were customer cash flow problems (51%), banking delays (26%), internal approval delays (23%) and complex payment processes (17%). The US report names liquidity first, followed by goods or services not delivered as promised, technical problems with online payment and slow internal approvals.

In practice, these headline reasons break down into a short list of causes:

  • Disputes. The customer disagrees with the price, quantity or delivery, and holds the whole invoice.
  • Missing purchase order or approval. The invoice cannot enter the customer’s payment run without a matching PO reference or a signed-off delivery note.
  • Wrong contact. The invoice or reminder goes to a buyer who has left, or to a general inbox no one reads.
  • No reminder at all. Nobody chases, so the invoice sits at the bottom of the customer’s queue.
  • Relationship fear. The supplier does not chase because it worries about the account. A Sage study of more than 3,000 small firms, now dated, found that over 30% had not chased a payment for this reason.

Only the first of these is a genuine disagreement. The rest are failures of information or process. That matters, because a reminder can fix an information failure. It cannot fix a dispute or an empty bank account.

What does automated follow-up demonstrably change?

Timely, personalised reminders raise payment rates and shorten time to payment. That finding is consistent across independent trials. We found no published randomised trial on commercial invoices specifically, so the best evidence comes from close analogues.

The closest is a field experiment by Gillitzer and Sinning with Australian Taxation Office business debtors, most of them micro-enterprises. A simple reminder letter raised the probability of payment by about 25 percentage points compared with no letter. Sending the reminder earlier produced faster payment, with no change in the ultimate payment rate. In plain terms, reminders bring money forward, and they recover some money that would not otherwise arrive.

Message design also matters. In a UK trial led by the Cabinet Office’s Behavioural Insights Team and summarised by J-PAL, 5,450 people with overdue court fines were randomly assigned text reminders or none. Texts raised the average amount paid from £4.46 to £10.94. A message that used the recipient’s name produced an average payment of £20.87, against £14.73 for the standard text. The setting is consumer, not commercial. The lesson carries over: a specific, addressed message outperforms a generic one.

Platform data points the same way, with less rigour. Intuit reports that AI-drafted reminders on overdue invoices get businesses paid an average of five days sooner, and that overdue invoices are 10% more likely to be paid in full. The comparison basis is not published, so treat this as platform-reported. Xero’s 2019 analysis of 10 million invoices, reported second-hand, found invoices with an online payment option were paid in about 12 days, against 22 without. That measures payment convenience, not reminders, which is itself a useful finding.

Vendor case studies show what is possible in favourable conditions. Versapay reports that one client cut DSO from nearly 50 days to about 30, and reduced the share of invoices more than 60 days overdue from 44.7% to 14.1%. Upflow reports that a Luxembourg manufacturer reduced invoices more than 180 days past due by 87% within two months. Both are vendor-reported and self-selected. Both bundled reminders with payment portals, system integration or formal notices, so the effect of reminders alone cannot be isolated.

What does automated follow-up not change?

Follow-up does not create cash a customer does not have, and it does not resolve a dispute on its own. Customer cash flow was the most cited cause of late payment in both the UK and US surveys. For those accounts, more reminders add cost and strain without adding recovery.

Three limits are worth stating plainly:

  1. Disputes need a person. A reminder on a disputed invoice that ignores the dispute damages the relationship the supplier was trying to protect.
  2. Bad data defeats automation. If the contact is wrong or the PO reference is missing, an automated sequence will send accurate reminders to the wrong place. The fix is upstream, in the order-to-invoice process. Our reconciliation work for a building materials distributor matched 97% of documents automatically, and clean matching is what makes follow-up land.
  3. The timing effect fades. The tax trial found that earlier reminders sped up payment, not that they raised the final payment rate. Once an account is well past due, the question changes from “when will they pay?” to “will they pay, and what will it cost to find out?”

Why does debt over 90 days need different handling?

Debt over 90 days is a different problem, not a later stage of the same one. The share of invoices this old is small, but the money at risk per invoice is high and the causes are harder. A reminder sequence designed for invoices five days late will not move them.

The Atradius UK data shows how past-due invoices are distributed. The table sets out that distribution with the handling each bucket typically needs.

Days past due Share of UK past-due invoices Typical cause What works
Under 30 73% Oversight, approval queue, payment run timing Automated, polite, personalised reminders with a payment link
31–60 14% Missing PO, wrong contact, unraised query Reminder plus a check of contact and documents; a call for larger balances
61–90 7% Dispute or early cash strain Named owner, dispute log, agreed payment plan
Over 90 6% Dispute, insolvency risk, lost contact Case-by-case triage: resolve, negotiate, escalate or provide for

Distribution: Atradius Payment Practices Barometer UK 2026, 210 UK firms. Causes and handling reflect common practice, not survey data.

For accounts in the oldest bucket, the useful question is economic. Which accounts are worth pursuing, by which route, and at what cost? Escalation through solicitors or the courts recovers money, but on the wrong case it costs more than it returns.

This is the problem we solved for a global credit insurer. We built a decision model, trained on the insurer’s own history of closed cases, that estimates for each overdue commercial account the likely recovery and cost of each route: amicable collection or legal action. It was back-tested on past cases before going live, and collection teams keep the final decision. The estimated result is more than €3M a year in avoided legal costs, mainly by not escalating cases that would have settled anyway. The same logic applies to a company prioritising its own aged debt: rank accounts by expected recovery after cost, and spend effort where it returns most. See our decision intelligence practice.

A note on scope. What we describe here is follow-up in the company’s own name, by its own finance team, supported by better data and prioritisation. It is not the hand-over of debt to a collection agency. Keeping follow-up in-house preserves the customer relationship, which is often the reason firms hesitated to chase in the first place.

How should you measure collections performance?

Measure four things monthly, against a baseline taken before any change: DSO, overdue share by ageing bucket, recovery rate on aged debt and write-offs. A single headline DSO figure hides too much.

Benchmarks help set ambition. The Hackett Group’s 2025 survey of the 1,000 largest US public companies found an 18-day DSO gap between top and median performers, worth $600bn in receivables. APQC benchmarks, older but widely cited, put top performers at 30 days or less and the median at about 36. Payment terms vary by sector, so compare against your own terms first.

How to measure it

  • DSO. Trade receivables divided by credit sales, multiplied by the days in the period. Track it monthly alongside your average payment terms, so a change in terms is not mistaken for a change in collection.
  • Overdue share by bucket. The value of receivables under 30, 31–60, 61–90 and over 90 days past due, each as a share of total receivables. Movement out of the older buckets is the clearest sign follow-up is working.
  • Recovery rate on aged debt. Cash collected on accounts that were over 90 days past due at the start of the period, as a share of their opening balance.
  • Cost to collect. Staff hours and external legal fees per pound or dollar recovered, especially on aged debt.
  • Write-offs. Bad debt written off as a share of credit sales, measured over a full year to smooth timing.
  • Dispute rate and cycle time. The share of overdue invoices with an open query, and the days taken to resolve one. This often explains more of DSO than reminder timing does.

Before starting, run the new process side by side with the current one on a defined group of accounts. That gives a fair comparison rather than a before-and-after story. We explain how we structure that kind of test in our note on where AI pays back in the back office, and in our finance operations practice.

What should a finance leader take from the evidence?

Automate the routine, and apply judgement to the exceptions. The independent evidence supports timely, personalised reminders as a reliable way to bring cash forward on the majority of overdue invoices, which are under 30 days late. It does not support sending more reminders to customers who are in dispute or short of cash.

The practical sequence follows from that. Fix contact and document data first. Automate polite, specific follow-up in your own name for young debt. Route disputes to a named owner quickly. For debt over 90 days, decide case by case on expected recovery after cost. Measure all of it by ageing bucket, not by DSO alone.

If you want to see what this would mean on your own numbers, our diagnostic shows it within seven days.

Sources

Frequently asked questions

Do automated payment reminders actually work for business customers?

Yes. In a randomised trial with Australian business tax debtors, a simple reminder letter raised the probability of payment by about 25 percentage points compared with no letter. Sending it earlier produced faster payment, though not a higher final payment rate.

How big is the late payment problem in the UK?

The Department for Business and Trade estimates late payment costs the UK economy almost £11bn a year and affects over 1.5 million businesses. Around £26bn is owed in late payments at any time, an average of £17,000 per affected business.

Why do B2B invoices go overdue?

UK firms surveyed by Atradius in 2026 most often cited customer cash flow problems (51%), banking delays (26%), internal approval delays (23%) and complex payment processes (17%). Many of these causes are administrative, which is why accurate, timely follow-up helps.

Should invoices over 90 days overdue be handled differently?

Yes. Older debt is more likely to involve a dispute, a missing approval or a customer in financial difficulty, so a standard reminder sequence has less effect. These accounts need triage: resolve the blocker, decide whether to escalate, and weigh the expected recovery against the cost of pursuing it.

What metrics should a finance team track for collections?

Track days sales outstanding (DSO), the share of receivables in each ageing bucket, the recovery rate on accounts past 90 days, and write-offs as a share of credit sales. Measure them monthly against a baseline taken before any change.

Does automated follow-up damage customer relationships?

Not when it is polite, accurate and sent in the company's own name. Personalisation matters: in a UK government trial, text reminders that addressed the recipient by name produced higher payments than a standard message.

Related insights

Article · Finance operations

Back-office AI ROI: which finance processes pay back first

Which back-office processes give AI a measurable return within months, how to size it from volume, cost and exceptions, and which to leave alone.

Guide · Approach

How to run a risk-free AI pilot that proves ROI before you commit

Most AI pilots never show a return. A six-stage method to de-risk one: a money metric, a backtest, pre-agreed thresholds, a control group and stop rules.

Start with a diagnostic.

Share one report. Within seven days, we show you in money what AI would have delivered on your own past numbers. If it does not deliver, you do not pay.

Book a diagnostic →