UK Payment Practices Reporting: What You Must Publish

A UK company or LLP that exceeded two or more of £54 million turnover, £27 million balance sheet total and 250 employees on its last two balance sheet dates is caught by the duty to report on payment practices and performance. It must publish, within 30 days of the end of each six-month reporting period, the average number of days taken to make payment measured from invoice receipt, the percentage of payments made in 30 days or fewer, in 31 to 60 days and in 61 days or longer, and the percentage of invoices not paid within the agreed period. The report goes on a public register under the company's own name.

Those threshold figures are the current ones. They apply for financial years beginning on or after 6 April 2025, and they replaced a £36 million turnover and £18 million balance sheet test that a considerable amount of third-party guidance still prints as though it were live. The uplift matters most to businesses sitting between the two sets of numbers. Some were released from the duty by the change; others are reading stale guidance and assuming they were.

From 1 January 2026 there is a second obligation on top of the first. Large companies must publish payment data in their directors' report for financial years beginning on or after that date. It does not replace the twice-yearly filing. It runs alongside it, over a different period shape, drawing on the same underlying invoice records, with a director's signature attached.

So payment practices reporting in 2026 means two publications built from one dataset, governed by definitions that were extended twice in 2025 and will be affected again by legislation now before Parliament.

The rules are not the hard part. They are published, specific and, on the points that matter, unambiguous. The hard part is that every statistic in the report is computed from fields most accounts payable ledgers were never designed to hold. The clock runs from the date an invoice was received, not the date on the invoice and not the date it was posted. The dispute metric added for financial years beginning on or after 1 January 2025 requires a per-invoice flag that no standard ledger carries. The construction retention metrics added from 1 April 2025 require data on money withheld in both directions, only one of which sits in payables at all.

That gap between the rules and the records is where payment performance reporting goes wrong. A business can read the guidance correctly, understand every definition, and still publish a figure it cannot defend, because the data it computed the figure from was the closest available approximation rather than the thing the regulations actually specify.

Who the Duty to Report on Payment Practices and Performance Applies To

The duty attaches to companies formed under the Companies Act 2006 and to limited liability partnerships formed under the Limited Liability Partnerships Act 2000. An entity is in scope for a financial year if it exceeded the medium-sized criteria on both of its last two balance sheet dates.

The payment practices reporting threshold, for financial years beginning on or after 6 April 2025, is a two-of-three test rather than a three-of-three test:

• turnover of more than £54 million
• balance sheet total of more than £27 million
• more than 250 employees

Exceeding any two of those on the last two balance sheet dates brings the entity into scope. A business with £60 million turnover, a £20 million balance sheet and 180 employees meets only one criterion and is out. A business with £60 million turnover, a £30 million balance sheet and 90 employees meets two and is in.

A parent company applies the same two-of-three test to the group's aggregate figures, with the two financial limits available in either form: turnover of more than £54 million net or £64 million gross, a balance sheet total of more than £27 million net or £32 million gross, or more than 250 employees across the group. Because the net and gross figures are alternatives, a group can qualify on gross figures alone even where consolidation adjustments bring the net position below the line.

Three exclusions apply. A business does not report in its first financial year. Overseas-incorporated companies are outside the duty entirely, including those with substantial UK trading operations. Entities not formed under the Companies Act 2006 or the LLP Act 2000 are outside it as well, which takes out most partnerships, charities constituted other than as companies, and public bodies.

Reporting periods

The normal pattern is two reporting periods per financial year. The first runs for the six calendar months from the first day of the financial year, and the second covers the remainder. Publication is due within 30 days of the end of each period.

Irregular financial years change the count. A financial year shorter than nine months requires a single report covering the whole year. A financial year longer than fifteen months requires three. Businesses changing their accounting reference date are the ones most likely to get this wrong, because the change usually happens for reasons unconnected with payment reporting and the filing consequence is easy to miss.

Which contracts count

Only qualifying contracts feed the statistics. A qualifying contract is one between two or more businesses, sufficiently connected with the UK, for goods, services or intangible property. Consumer contracts are out. Employment is out. Financial services contracts are specifically excluded, and the same carve-out applies to the directors'-report calculations, so a business does not need two different views of which contracts count.

The government publishes a scope checker for the duty to report alongside its guidance, and for a marginal case that is the reference to work from rather than any secondary summary. Scope is the one part of this regime where the authoritative source is both free and definitive.

Groups operating internationally should treat scope as a per-jurisdiction question rather than a group-wide one. Australia's Payment Times Reporting Scheme imposes a comparable obligation on large Australian entities, but its thresholds, period definitions, metric formulas and filing portal are its own. An entity in scope in both places produces two reports from two differently-defined datasets, and neither set of figures can be derived from the other.

Two Reporting Surfaces Drawing on One Dataset

Until recently, a business in scope had one place to publish. It now has two, and most of the available commentary covers one or the other rather than both. Corporate governance advisers write about the directors' report and ignore the portal. Accounts payable and procurement content covers the portal and predates the directors'-report duty entirely.

The first surface is the twice-yearly portal filing. Reports are submitted through the government's web service and appear on a public register that anyone can search by company name. This is the long-standing obligation, created by the Reporting on Payment Practices and Performance Regulations 2017 and running since April that year.

The second surface is the directors' report. The Companies (Directors' Report) (Payment Reporting) Regulations 2025 came into force on 1 January 2026 and require large companies to publish payment data within the directors' report for financial years beginning on or after that date. The figures sit inside the annual accounts, subject to the ordinary approval and signature process that applies to everything else in that document.

The two duties do not catch identical populations. The directors'-report requirement applies to large companies only, not to small or medium-sized ones, and only from a company's second financial year onward. A parent company reports on the consolidated group. Subsidiaries that are small, medium-sized, or in their first financial year can be excluded from the group figures.

What the directors' report has to contain

• narrative on the company's standard payment terms and any changes made to them during the period
• the average time taken to make payments in the reporting period
• the percentage of payments made in 30 days or fewer, in 31 to 60 days, and in 61 days or longer
• the monetary sum of payments falling in each of those three bands
• the sum of payments due but unpaid beyond agreed terms
• the proportion of payments not made within agreed periods

The monetary-value data points are the newest element, and they commence with the rest of the directors'-report duty for financial years beginning on or after 1 January 2026. That is a full year later than the portal additions, which attach to financial years beginning on or after 1 January 2025. The two dates are easy to conflate, and conflating them puts a company a year out on which figures it owes.

Financial services contracts are excluded from the directors'-report calculations, matching the portal report's qualifying-contract carve-out.

The annual figure is not the average of the two halves

The definitions are shared across both surfaces, which is the useful part. The government guidance on reporting payment data in directors' reports restates the payment clock in the same terms the portal guidance uses: Day 1 is the day after the date on which the company receives an invoice or other notice of the amount to pay, with the period ending when the supplier receives the funds. The size test is the same too, exceeding two or more of £54 million turnover, £27 million balance sheet total and 250 employees. One correctly-built dataset therefore serves both publications.

What does not carry across is the arithmetic. The directors' report covers a full financial year; the portal covers two six-month windows. The annual average days to pay is not the mean of the two half-year averages, because the statistic is weighted by invoice count rather than by period. A business that pays 2,000 invoices in an unusually slow first half and 8,000 invoices quickly in the second half will produce an annual figure much closer to the second half than a midpoint would suggest.

The consequence is practical. The full-year figure has to be recomputed across the whole invoice population rather than derived from the two figures already filed. A team that discards or overwrites the underlying invoice-level records after each portal filing will find itself rebuilding them at year end, under accounts-preparation deadlines, for a document a director signs.

What Must Be Published in Every Reporting Period

The payment practices reporting requirements fall into three layers: the statistics, the narrative statements, and a short set of yes-or-no declarations. Three statistics are required in every reporting period, regardless of financial year:

• the average number of days taken to make payment, measured from invoice receipt
• the percentage of payments made in 30 days or fewer, in 31 to 60 days, and in 61 days or longer
• the percentage of invoices not paid within the agreed period

Two more were added for financial years beginning on or after 1 January 2025: the sum total of payments made during the reporting period, and the percentage of payments not made within agreed terms because of a dispute.

A further set was added for financial years beginning on or after 1 April 2025, covering the use of retention clauses in qualifying construction contracts.

Working out which metrics apply to you

Every commencement date in this regime attaches to the start of the financial year, not to the reporting period and not to the filing date. That distinction is where the guidance is most often misread.

A company with a financial year running from 1 October 2024 to 30 September 2025 files two reports during 2025, and neither includes the dispute percentage or the payments total, because the financial year began before 1 January 2025. The same company's next financial year begins on 1 October 2025, which falls after both the 1 January and the 1 April commencement dates, so both of its reports carry the FY-2025 additions and, if it has qualifying construction contracts, the retention metrics as well. The directors'-report duty still does not apply, because that financial year began before 1 January 2026. Its financial year beginning 1 October 2026 is the first to fall under every tranche.

A company on a calendar financial year picks the additions up in a different order. FY 2025 takes the payments total and the dispute percentage, but not construction retention, because that year began before 1 April. FY 2026 then takes the retention metrics and the directors'-report duty with its monetary bands together.

The narrative statements

The statistics are published alongside written statements, and both parts sit on the same public record. They are read together, which is worth remembering when the narrative describes terms the figures do not appear to support.

A report must set out the business's standard payment terms, any changes made to those terms during the reporting period, and how suppliers were notified of the change. It must also describe the process for resolving disputes related to payment. This is a description of what actually happens rather than a policy statement, and it is the section a supplier or journalist reads first when the figures look poor.

The declarations

Four yes-or-no declarations complete the report:

• whether e-invoicing is offered to suppliers
• whether supply chain finance is available to suppliers
• whether the business is a member of a payment code, and which one
• whether the business deducts sums from invoices as a charge for remaining on a supplier list

The last of these is the one that carries reputational weight. Pay-to-stay arrangements are legal but poorly regarded, and the declaration exists to make them visible.

The e-invoicing declaration has become more consequential than it looks. A business answering it in 2026 is not only describing current practice but implicitly signalling its position relative to the UK e-invoicing mandate timetable, and the answer is likely to move from no to yes for most reporting businesses over the next few filing cycles.

The Clock Starts on Invoice Receipt, Not the Invoice Date

Day 1 is the day after the date on which the business receives the invoice, or other notice of the amount to pay. The period ends on the day the supplier receives the payment. Everything else in the calculation follows from those two points.

Neither of them is a field most accounts payable systems reliably hold.

The invoice date is whatever the supplier typed on the document. It can precede receipt by a day or by three weeks, depending on the supplier's own billing cycle, postal arrangements and month-end habits. Using it as the start point flatters the business on invoices that arrive late and penalises it on invoices that arrive promptly, in neither case producing the number the regulations ask for.

The posting date is the date somebody keyed the invoice into the ledger. It sits after receipt by however long the invoice waited in an inbox, on a desk, or in an approval queue before anyone recorded it. Using it as the start point deletes exactly the delay the statistic was written to expose. A business with a three-week intake backlog and a five-day payment run reports five days.

The gap between those two dates is not a rounding difference. In most AP functions it is the largest single component of the published average.

The emailed PDF problem

For invoices arriving as attachments into a shared accounts payable mailbox, the receipt date exists in exactly one place: the timestamp on the email, and only until someone deletes it or the mailbox retention policy does. Once the invoice has been detached, approved and posted, the ledger holds a posting date and an invoice date, and the receipt date is gone.

It cannot be recovered later. It can be estimated, and this is where businesses get into difficulty, because an estimate published as a statutory figure is a representation about data the business does not have. Publishing information that is false or misleading, knowingly or recklessly, is a separate offence from failing to publish at all. A team that reconstructs receipt dates from posting dates and files the result has produced a number it cannot substantiate if anyone asks how it was derived.

The end of the period has a similar trap at the other end. The clock stops when the supplier receives the funds, not when the payment was authorised, not when the payment run was approved, and not when the BACS file was submitted. For a business making payments on standard three-day BACS cycles, treating the release date as the end date understates every payment in the population.

What the dataset has to contain

Before any of the required statistics can be computed, a business needs a per-invoice record, covering every qualifying invoice in the reporting period, holding at minimum:

• the supplier identity
• the date the invoice was received
• the agreed payment period for that contract
• the invoice value
• the date the supplier received payment
• for financial years beginning on or after 1 January 2025, whether the invoice was disputed

Not a sample. Not a summary by supplier. The average is invoice-weighted, the band percentages are counts across the population, and the Fair Payment Code tests are distributions, so all three need the full set of rows.

Two of those fields are properties of the invoice and its contract: the receipt date and the agreed payment period, alongside the invoice value. Where invoices arrive as PDFs rather than through a structured channel, those fields have to be read off the documents themselves, which is the practical constraint most teams hit first. It is possible to extract invoice-level data from supplier PDFs into a spreadsheet at population scale, with the fields defined once in a prompt so the same columns come out for every supplier format in the batch, and with each output row carrying a reference back to the source file and page so a published figure can be traced to the document behind it.

The boundary is worth stating plainly, because it is the part vendors tend to blur. Extraction builds the invoice half of the dataset. The payment date comes from the ledger or the bank, not from the invoice, and the receipt date has to be captured at intake rather than derived from the document, since an invoice never states when it arrived. The report is assembled from both halves, and no extraction step produces it on its own.

This is not days payable outstanding

Finance teams already track payment speed, and the instinct is to reach for the existing measure. It does not transfer. How days payable outstanding is calculated is a matter of internal choice: it works from balances and cost of sales, it is computed over whatever period suits the analysis, and reasonable people compute it differently.

The published average time taken to pay suppliers is none of those things. It has a prescribed start point, a prescribed end point, a defined population of qualifying contracts, and a fixed period. It is published under the company's name on a register that anyone can search, and misstating it carries criminal liability. A DPO figure and a statutory average calculated over the same period will not match, and should not be expected to.

Disputed Invoices Still Count as Late

For financial years beginning on or after 1 January 2025, reports must include the percentage of payments not made within agreed terms because of a dispute.

A disputed invoice paid outside its agreed terms is still a late payment, and it still appears in the percentage of invoices not paid within the agreed period. The dispute metric sits beside that figure and qualifies it. It does not net anything out of it. Several widely-circulated summaries of the 2025 changes leave this ambiguous, and a finance team that reads the new metric as an exemption will build its dataset around the wrong assumption.

The policy reasoning makes the design obvious once stated. The metric exists to make visible the practice of raising a dispute in order to delay payment. A figure that removed disputed invoices from the late count would reward precisely the behaviour it was introduced to expose. Instead the reader of a published report sees both numbers and can draw their own conclusion about a business whose late-payment percentage is high and whose disputed proportion is higher still.

The distinction no ledger draws

The metric requires a business to separate, invoice by invoice, two reasons for lateness that most systems record identically:

• late because the invoice was genuinely disputed on price, quantity, quality or contractual entitlement
• late because it sat in an approval queue, was missing a purchase order reference, or was waiting on someone who was away

Both typically appear in the ledger as an unapproved invoice, or as a hold with a free-text reason that was written for internal workflow purposes rather than statutory reporting. Hold reason codes, where they exist at all, tend to have grown organically and to mix genuine commercial disputes with routine processing exceptions under the same label. Neither an approval status nor a hold code maps cleanly onto the statutory distinction, which is why this metric is more work than its single line in the guidance suggests.

It cannot be applied retrospectively with any credibility

A dispute is a state that existed at a particular time, evidenced by a query raised with the supplier, a credit note requested, a delivery rejected, a contract clause invoked. Deciding at reporting time which of the past six months' late invoices were "really" disputed is not a data-cleaning exercise. It is a judgement made by a party with an interest in the answer, about events for which contemporaneous evidence either exists or does not.

That has a straightforward operational implication. The dispute flag has to be captured as part of handling the invoice, from the first day of the financial year the metric applies to. A business whose relevant financial year has already started and which has not been capturing it has three realistic options: recover the position from contemporaneous evidence such as supplier correspondence where that evidence genuinely exists, report a conservative figure and be able to explain the basis for it, or report zero disputes for the period. The third is defensible in a way that an invented percentage is not.

The percentage of invoices not paid within the agreed period has been required since the duty began. The dispute percentage now runs alongside it, and the directors' report adds the sum of payments due but unpaid beyond agreed terms, putting a monetary value on the same population. Three figures, three angles on the same set of invoices, and each one wrong if the underlying per-invoice records are wrong.

Construction Retention Reporting Needs Data in Both Directions

For financial years beginning on or after 1 April 2025, a reporting business with qualifying construction contracts must report on its use of retention clauses. This applies whether or not the business considers itself a construction company. What matters is holding qualifying construction contracts, which catches retailers, hoteliers, landlords and manufacturers running fit-out or capital works programmes.

The first disclosure is a yes or no: does the business use retention clauses in its construction contracts. A no ends the section. A yes opens a set of further disclosures:

• the pattern of use, meaning whether retention is applied to all construction contracts or only to some
• the contract-sum threshold above which retention is applied
• the standard retention percentage
• the mechanism by which retentions are released

Then two performance metrics, and these run in opposite directions. One covers retentions the reporting business withholds from its own suppliers and subcontractors. The other covers retentions withheld from the reporting business by its customers.

Why the second direction is the hard one

Money the business withholds from suppliers is at least adjacent to systems the finance team controls. It shows up as a deduction against subcontractor applications, it reduces the certified payment, and someone in the business decided to apply it.

Money withheld from the business by a customer is not in the payables ledger at all, because it is not a payable. It is a reduction against amounts the business is owed, sitting in receivables, in payment certificates issued by somebody else's quantity surveyor, and in the contract documents that set the percentage and the release conditions. In many organisations the only complete picture lives with the commercial or project team rather than with finance, held in a contract register or a spreadsheet per project.

Retention data also has a different shape from everything else in the report. Every other figure is driven by invoice dates and payment dates against agreed terms. Retention release is governed by contractual milestones: practical completion, the end of the defects liability period, the issue of a making-good certificate. Those dates are not in any invoice, they vary by contract, and they routinely slip. A business that tries to force retention into the same date arithmetic as the payment statistics will produce something that does not reconcile.

The usual finding, for a business reporting on this for the first time, is not that the data is missing. It is that the data exists in project-level records that were never designed to aggregate. The discipline needed here is the same one that underpins tracking subcontractor retention release across construction projects: a single register holding, per contract, the retention percentage, the sum withheld to date, the release trigger and the date of each release, in both directions. Assembled once, it answers the statutory disclosures directly. Assembled from scratch every six months, it consumes a fortnight and produces figures nobody is confident in.

The Fair Payment Code and Its Gold, Silver and Bronze Tiers

The Fair Payment Code is administered by the Office of the Small Business Commissioner and awards businesses one of three tiers based on how quickly they actually pay.

Gold requires paying at least 95% of all invoices within 30 days.

Silver requires paying at least 95% of all invoices within 60 days, including at least 95% of invoices from small businesses within 30 days.

Bronze requires paying at least 95% of all invoices within 60 days.

Awards last two years, after which a business must reapply and demonstrate that it still meets the standard. There is no charge to apply.

As of August 2026 the Code had 713 awardees listed on the Small Business Commissioner's Fair Payment Code register, of which 316 held Gold. That number has moved quickly since launch, so check the register rather than relying on any figure quoted in secondary coverage, including this one.

It replaced the Prompt Payment Code

The Fair Payment Code launched in December 2024 and replaced the Prompt Payment Code. The Prompt Payment Code is no longer the current scheme, and a business describing itself as a signatory to it is describing a historic status rather than a live one. Where a payment practices report asks whether the business is a member of a payment code, the Fair Payment Code is the answer that means something now.

The change was not only a rebrand. The Prompt Payment Code operated as a set of commitments a business signed up to; the Fair Payment Code is an award granted against a measured threshold, verified before it is issued, and time-limited. The tiers exist to be climbed, and the structure assumes businesses enter at Bronze and work upward rather than treating their first award as a permanent classification.

The tests read against the same dataset as the statutory report

This is the part that connects the Code to everything else a reporting business is already doing, and the part most often misunderstood.

The qualifying tests are distributions, not averages. A business cannot infer its tier from its published average days to pay, because an average of 24 days is entirely compatible with 12% of invoices taking more than 30 days, and 12% fails the Gold test. The question the Code asks is what proportion of the invoice population fell inside each window, which is a count across every invoice rather than a single summary figure.

That is the same population, computed from the same per-invoice records, that the statutory report requires. Run the distribution once and the answer is definitive before an application is ever submitted.

One additional field is needed for Silver that the statutory report does not require. The Silver test distinguishes invoices from small businesses, which means knowing which suppliers are small. That is an attribute of the supplier master rather than of any invoice, and it usually has to be established separately, whether from Companies House filings or from the suppliers themselves.

How to Publish, and the Penalties for Getting It Wrong

Reports are published through the government's payment practices reporting service, which requires a Companies House Online account. Submission is per reporting period rather than annual, so a business on the normal pattern files twice a year, each time within 30 days of the period ending. There is no fee and no filing extension mechanism.

Published reports go onto a public register that anyone can search by company name. They stay there. A supplier deciding whether to take on a new customer, a journalist writing about payment terms in a sector, and a competitor benchmarking its own figures all have access to the same history, and the sequence of a company's reports over time is as visible as any single one of them.

Two separate offences

Failing to publish a report is a criminal offence, and it is committed by the business and by every director, or in the case of an LLP by every designated member. Personal liability is the design, not an edge case.

Knowingly or recklessly publishing information that is false or misleading is a separate offence. It stands on its own, which means a business that files on time can still be liable on the basis of what it filed. Both offences are punishable on summary conviction by a fine.

A defence is available to a director who took all reasonable steps to secure compliance.

That defence is where the shape of a business's underlying records starts to matter legally rather than just operationally. A director asked to demonstrate reasonable steps is in a stronger position where the published figures were computed from invoice-level records that can be produced and reperformed, and a considerably weaker one where the figures were derived from ledger fields chosen because they were available rather than because they matched the statutory definitions. The word "recklessly" does real work in the second offence: it does not require an intention to mislead, only publication without proper regard for whether the figures were right.

The directors'-report figures sit under a different liability regime again. They form part of the annual accounts and carry the responsibilities that attach to the directors' report as a whole, which run alongside the portal offences rather than duplicating them. A business in scope for both is exposed on two fronts from one set of numbers.

What the Commercial Payments Bill Would Change

The Commercial Payments Bill, also known as the Small Business Protections (Late Payments) Bill, had its first reading in the House of Lords on 19 May 2026 and is at committee stage. Subject to Royal Assent, it is expected to come into force in 2027.

It would change the substance of what businesses report on rather than the reporting mechanics.

Maximum payment periods. 30 days for the public sector and 60 days for the private sector, with any longer contractual term automatically void. A void term does not simply become unenforceable at the supplier's option; it ceases to be the agreed period, which matters directly for a statistic defined by reference to agreed terms. A proposed further reduction to 45 days after five years was trailed and has been dropped following consultation.

Mandatory interest. Interest on late payments at 8% above the Bank of England base rate.

A fixed sum for late or unsupported disputes. Suppliers would gain a right to a fixed payment where a purchaser raises a dispute late, or without providing sufficient information to justify it. This is aimed at the same behaviour the FY-2025 dispute metric was written to expose, approached from the other end: the metric makes tactical disputes visible, and the Bill attaches a cost to them.

A retention ban, phased. Retentions in construction contracts would be prohibited, but through a two-year transition of phased reduction rather than an overnight prohibition.

The tension in the retention rules

Construction retention reporting arrived for financial years beginning on or after 1 April 2025. The Bill would abolish the practice those metrics measure. Because the prohibition comes through a phased transition rather than a single commencement date, retention reporting and the retention ban would coexist for a period, with businesses reporting on a declining retention percentage they are contractually required to reduce.

Anyone building retention reporting infrastructure now should build it knowing its useful life is measured in a handful of reporting cycles, and that during the transition the figures will move for regulatory reasons rather than commercial ones.

What the interest provision actually changes

A UK supplier is already entitled to charge interest on a commercial debt paid late, at 8% above base rate, under existing legislation. What the Bill changes is not the rate but the exercise of the right. The current position leaves it to the supplier to invoke, and in practice most do not, because a small supplier that charges statutory interest to a large customer is making a commercial decision about the relationship as much as a legal one. If interest becomes mandatory rather than elective, that calculation disappears, and late payment acquires an automatic and quantifiable cost.

The gap between what the law permits and what suppliers actually do is wide enough that many purchasers have never encountered it. A business modelling its exposure should start from statutory interest a supplier can charge on a late UK invoice applied across its own published late-payment percentage, because that percentage is already public and is the closest available proxy for what mandatory interest would cost.

Every provision in the Bill is measured against the same fields the current report already requires: when the invoice was received, what period was agreed, what it was worth, when the supplier was paid, and whether it was disputed. The 60-day cap is a test against agreed terms. Mandatory interest is a calculation over the interval between due date and payment date. The fixed-sum right turns on when a dispute was raised.