You have been told to put a late fee in your terms, switch on automatic reminders, offer a discount for early payment and invoice promptly. We tested all of it against 61,185 real invoices. The thing almost every owner believes about large invoices turns out not to be true, the late-fee field has never once been filled in, and reminders are set on one invoice in thirty.
Key figures
61,185 invoices from 271 businesses whose invoices met our inclusion rules, December 2007 to October 2026. Free to reuse under CC BY 4.0 with attribution. Download the CSV
If you have ever read an article about getting paid on time, you have read the same four suggestions. Put a late fee in your terms so clients take the due date seriously. Switch on automatic reminders so you are not the one chasing. Offer a small discount for paying early. Get the invoice out quickly. All four sound obviously right, and none of the articles recommending them can tell you whether a single small business actually does any of it.
We can, for the parts that leave a trace in invoicing software. Some of what follows is a finding. Some of it is a question we decided we could not answer honestly, and those are marked.
Almost every owner treats the largest outstanding invoice as the one at risk. Across undiscounted US dollar invoices, mean days to pay runs 17.5 days under $500, 20.1 days from $500 to $2,000, 17.5 days from $2,000 to $10,000 and 18.6 days over $10,000. Four bands, between 56 and 103 businesses each, and the $10,000 invoice is paid about as quickly as the $400 one.
This is the best-supported comparison in the study and it is worth acting on, because it tells you that invoice size is a poor guide to which payment to worry about. If you are going to spend attention on one invoice, the useful signals are elsewhere: whether it has gone past due at all, and whether the first payment arrived in full.
One limit worth stating: these are invoices with no discount applied, which is the comparison group from section 5. We cannot report the discounted equivalent, for the reason given there.
Billbooks has a late-fee field. You can set it as a percentage or a flat amount, and it appears on the invoice and can be charged automatically. Across 68,337 invoices from 402 businesses, paid and unpaid, drafts included, the number with that field filled in is zero, and the number with a fee actually charged is also zero.
Be careful what that does and does not mean. It is a measurement of one structured field, not of the practice. Owners may well have late-fee wording in their written terms, and this study does not read free-text fields, so it cannot say how many do. What it shows is that nobody uses the mechanism that would actually apply a fee. The simplest explanation is not that owners have weighed late fees and rejected them: it is that a field nobody goes looking for is a field nobody fills in, and a fee you never charge is not a policy either way.
If you are considering one, the useful first step is working out what you would actually charge on a real overdue invoice, because the number is usually smaller than people expect. Our late fee calculator does that arithmetic.
Automatic reminders are the second piece of standard advice, and the setting is an interval in days rather than a switch: remind every 7, 15, 30 or 45 days after the due date. The interval is attached to 2,026 of 61,185 invoices, or 3.3%, and about a quarter of the businesses in the study have ever set one. The most common choice is weekly.
Where an interval is set, a reminder is often never due. A reminder fires a whole interval after the due date, so an invoice paid before then produces none: of the invoices carrying an interval, 589 were paid by the due date and most of the rest were paid inside their own interval. The group where a reminder was genuinely due and did not send falls below our reporting threshold in every bucket, which is the honest way to say it is small.
We are not publishing whether reminders get invoices paid faster, and that is a deliberate refusal rather than a gap. Reminders are not assigned at random: an owner switches them on for the client they already expect to chase, so the invoices that carry reminders are selected for being slow and any comparison between the two groups measures that selection. Our payment study declined the same comparison in August for the same reason, and declining it again is the consistent answer rather than a convenient one. We also looked at it per interval, and the apparent effect sits entirely in one bucket while the others show none, which is what a real effect does not look like.
Most invoices are paid once and closed: 50,905 of them, settling 26.0 days after the invoice date. But 2,846 invoices from 77 businesses took more than one payment, and those settle in 37.3 days, about 40% longer. Roughly one invoice in twenty. The tail is slower still: the 417 invoices settled in three to five payments took 63.3 days.
If you track only whether an invoice is paid or unpaid, a part-paid invoice looks like progress. On this evidence it is better read as a signal that the rest of that money is on a longer clock than usual.
We computed it and we are not reporting the figure. The raw comparison runs against the discount, and written up without a second thought it would be a confident headline saying early-payment discounts buy you nothing.
A discounted invoice is usually a bigger, negotiated job, so the only fair test compares discounted against undiscounted invoices of similar size. Doing that splits the data into eight groups, and all four discounted groups fall below our reporting threshold of 25 businesses: 15, 17, 20 and 7 businesses. A comparison is only as publishable as its weakest group. So the controlled comparison does not exist, the raw one cannot be told apart from invoice size, and the answer is that we do not know.
We are not printing the raw number at all, because a figure that cannot be interpreted is a figure that will be misread, and this one would have been the most quotable thing in the study. If you are weighing a discount, the question worth answering is what the discount costs you against what the wait costs you, which our early payment discount calculator works out for your own numbers.
For invoices whose lines carry their own work dates, we can see the gap between the work and the invoice. Of 13,803 such invoices, 6,606 are dated the same day as the earliest work on them or earlier, 4,170 go out within a week, 2,011 within a month and 1,016 more than a month after. So 21.9% leave more than a week after the job and 7.4% more than a month after. Drafts, voided invoices and recurring templates are excluded, because none of them ever went out.
This is the one piece of standard advice the data shows people broadly following, and it is the cheapest to act on. A month of delay before you invoice is a month added to every payment term you then set, which our due date calculator makes concrete.
Aggregate queries over Billbooks invoice records, taken from a single snapshot so no two figures here can disagree because the data moved between them. Sample and test data excluded. Negative days to pay, and days to pay over 365, are treated as data errors and dropped; that bound is not applied to settlement time, because an invoice whose last instalment lands more than a year out is a slow payment plan rather than an error. Days to pay is measured to the first payment recorded against an invoice, and days to settle to the last.
Every figure here comes from a bucket holding at least 25 distinct businesses and at least 100 documents. Buckets below that are suppressed rather than rounded, and the full list of what was suppressed ships with the data, because a bucket too small to report is sometimes itself the point. No organisation, client, person, document or free-text field is read or reported: the queries select record identifiers, dates, statuses, numeric settings and totals, and nothing else, so nothing identifying can reach this page.
The run is read only: a read-only transaction, aggregate selects, no writes. Our two earlier studies were run against a restored copy of the database and this one ran against the live system under those constraints, which is the convention our internal reporting uses. Aggregate, anonymised research of this kind is covered by the terms customers agree to; our privacy policy sets out how account data is handled.
The script that produces these numbers is committed, so a re-run moves them. Figures measured 22 October 2026. Companion studies: what happens after an invoice goes out and what happens to a quote. Reuse any of this under CC BY 4.0 with a link, or take the CSV.
Source: Billbooks payment-advice study, 61,185 invoices from 271 small businesses, December 2007 to October 2026 (aggregate only, figures shown only where 25+ businesses and 100+ documents contribute).
Barely. Across undiscounted US dollar invoices, mean days to pay is 17.5 under 500 dollars, 20.1 from 500 to 2,000, 17.5 from 2,000 to 10,000 and 18.6 above 10,000. A ten thousand dollar invoice is paid about as quickly as a four hundred dollar one, so invoice size is a poor guide to which payment to worry about.
Never, in this data. Across 68,337 Billbooks invoices from 402 small businesses, the structured late-fee field has never been filled in and no fee has ever been charged. This measures the field rather than the practice: an owner may still have late-fee wording in their written terms, which this study does not read.
About 40% longer. An invoice settled in more than one payment clears 37.3 days after the invoice date against 26.0 days for one paid in a single payment, across 2,846 invoices from 77 businesses. Roughly one invoice in twenty arrives in pieces, and the slowest tail runs to 63.3 days.
About 3.3%, or 2,026 of 61,185 invoices, set by roughly a quarter of the businesses in this study. The setting is an interval in days rather than an on and off switch, and the most common choice is a reminder every week after the due date.
This study will not say, and the refusal is deliberate. Reminders are not assigned at random: an owner switches them on for the client they already expect to chase, so invoices carrying reminders are selected for being slow and any comparison measures that selection rather than the reminder. Looked at per interval, the apparent effect sits in one group and the others show none.
This study does not know. Comparing like with like by invoice size puts all four discounted groups below our 25-business reporting threshold, so the controlled comparison does not exist. The raw comparison cannot be told apart from invoice size, so we do not report it.
Mostly straight away. Of 13,803 invoices whose lines carry their own work dates, 6,606 are dated the same day as the earliest work on them or earlier. About 21.9% go out more than a week after the job and 7.4% more than a month after, and a month of delay is a month added to every payment term you then set.