PayslipIQ Data — Modelled Study
A transparent, reproducible model of where UK PAYE payslips most commonly go wrong, what those errors cost, and how to spot them on the page you were just handed.
By Rocco Clayfield, Founder and Editor · Published 17 May 2026 · Updated 5 August 2026 · Jump to methodology
11.4% of simulated UK PAYE payslips carried at least one calculation discrepancy in our model.
Across 10,000 synthetic payslips generated against HMRC 2026/27 tax tables and weighted to the ONS Annual Survey of Hours and Earnings distribution, more than one in nine carried a simulated tax, National Insurance, pension or deduction error. Most were small. A non-trivial minority cost the modelled employee £500 or more across the year. This is a modelled estimate, not an observed measurement — the headline rate follows directly from the error-injection assumptions listed in the methodology below.
Tax codes drift after every job change
In our model, 18.2% of payslips in the first three pay periods following a job change carried a BR or 0T emergency code that had not yet rolled forward. The rate is a PayslipIQ modelling assumption: CIPP's report Systemic issues in HMRC RTI data collection (28 February 2025) documents this failure mode through employer case studies, but publishes no rate.
Salary sacrifice frequently miscoded
In 9.8% of modelled salary-sacrifice cases the sacrificed amount failed to reduce gross taxable pay on the payslip, costing the employee both income-tax and NI relief. HMRC manual EIM42750 defines how a valid sacrifice must operate; the 9.8% rate is a PayslipIQ modelling assumption.
Student-loan plan confusion is the second-biggest error
In our model, 13.7% of payslips with a student-loan deduction used the wrong plan, most commonly Plan 1 instead of Plan 2. The rate is a PayslipIQ modelling assumption; the thresholds are real. For 2026/27, GOV.UK publishes Plan 1 at £26,900 and Plan 2 at £29,385 (gov.uk/repaying-your-student-loan) — a gap wide enough to materially over- or under-deduct.
NI category letters get stuck at age boundaries
In our model, 11.4% of NI errors clustered at the 21st, 25th and State Pension age boundaries where category letters should change automatically (e.g. M to A, A to C). The correct letter for each circumstance is published at gov.uk/national-insurance-rates-letters; the 11.4% rate is a PayslipIQ modelling assumption.
Roughly half of eligible couples never claim Marriage Allowance
HMRC has estimated around 4 million couples are eligible for the Marriage Allowance (worth up to £252 a year), with more than 1 million yet to claim at the time of its June 2018 press release; the Low Incomes Tax Reform Group maintains current guidance on claiming. In our model, 7.6% of eligible couples in the sample had no transfer recorded against the lower earner — a modelling assumption, not an HMRC uptake statistic.
When an employee starts a new job before HMRC processes the P45 from the previous one, the new employer applies an emergency BR (Basic Rate, 20% on every pound) or 0T (no allowances) code. The Real Time Information system usually corrects this within one or two cycles, but in 18.2% of modelled job changes the wrong code persists for three pay periods or more. The fix is HMRC's Personal Tax Account or a call to 0300 200 3300; refunds typically flow through the next payroll cycle. See our tax codes guide.
English Plan 2, Scottish Plan 4, Plan 1 (pre-2012 English/NI), the new Plan 5 (English students from August 2023) and the Postgraduate Loan all have different thresholds and recovery rates. GOV.UK publishes the table annually; for 2026/27 the Plan 2 threshold is £29,385 while Plan 1 is £26,900. New employees who tick the wrong box on the Starter Checklist are over-deducted at 9% above the lower threshold. The fix is an SL2 stop notice from HMRC. See our student loans guide.
National Insurance category letters (A, H, M, V, C, X and the rest) determine the rate at which both employee and employer pay NI. They should change automatically on the employee's 21st birthday (apprentice levy), 25th birthday (under-25 reliefs) and State Pension age (employee NI ceases). In 11.4% of modelled boundary events the letter did not flip in the period it should have. GOV.UK's National Insurance rates and category letters page sets out the correct letter for every combination.
A genuine salary sacrifice arrangement reduces gross pay before income tax and NI are calculated. If payroll software treats the contribution as a post-tax deduction instead, the employee loses both income-tax and NI relief and the employer loses Class 1 secondary NI relief. The Optional Remuneration Arrangements rules from April 2017 narrowed which benefits qualify, but pensions, cycle-to-work and ultra-low-emission cars are still in. See HMRC manual EIM42750.
PAYE works on a cumulative basis, projecting your monthly pay over the rest of the tax year. A one-off bonus paid in month 3 can push the projection into the 40% band, triggering temporary over-deduction. The system self-corrects across subsequent months provided the cumulative code is operated correctly; an employee on a Week 1/Month 1 emergency code will not see that correction until year-end via a P800.
Marriage Allowance lets a non-taxpayer transfer 10% of their personal allowance (£1,260 in 2026/27) to a basic-rate-paying spouse, saving the couple up to £252 a year. HMRC has estimated around 4 million couples qualify, with more than 1 million yet to claim as of its June 2018 campaign. The transfer appears on the higher earner's payslip as a suffix-M tax code; on the lower earner's as suffix N. Backdating up to four tax years (to 2022/23, as of the 2026/27 tax year) is possible. Claim at gov.uk/marriage-allowance.
Employer childcare voucher schemes closed to new entrants in October 2018, replaced by Tax-Free Childcare. Existing members can continue, but only while they remain with the same employer and on a continuous voucher election. In 6.1% of modelled relevant cases, voucher deductions persisted after a job change or a 52-week break, breaching the rules and creating a taxable benefit-in-kind exposure.
UK case law since Bear Scotland v Fulton (2014) and Harpur Trust v Brazel (2022) has steadily expanded what counts as "normal remuneration" for holiday pay: regular overtime, commission and shift premia all in scope. Payroll systems that pay holiday at basic rate only systematically underpay. The Employment Rights (Amendment, Revocation and Transitional Provision) Regulations 2023 reinstated 12.07% rolled-up holiday pay for irregular-hours workers from April 2024.
Workers must be auto-enrolled into a qualifying pension on the pay reference period containing their 22nd birthday, provided they earn above the £10,000 trigger. In 5.1% of modelled 22nd-birthday events, enrolment did not happen until the following payroll month, costing the employee one period's contribution and the employer's 3% match. The Pensions Regulator publishes monthly compliance and enforcement bulletins.
An employee with a second job almost always has their full personal allowance applied at the main job and a BR code on the second. But for higher earners whose main job uses the full basic-rate band, the second job should be coded D0 (40% on every pound) or D1 (45%) instead of BR. In our model, 4.8% of dual-employment payslips carried the wrong second-job code, with HMRC reconciling later via a P800.
A modelled 32-year-old earning £42,000 in Manchester moved jobs in May. Their previous employer's P45 was not processed by HMRC for nine weeks, so the new employer applied a BR code on day one. Over seven pay periods, £6,300 of gross pay was taxed at a flat 20% rather than benefiting from the £12,570 personal allowance and the basic-rate band; the result was £1,089 of additional tax against the cumulative position they should have been in.
How they'd spot it on the payslip: the "Tax code" field reads BR with no number, and the "Taxable pay to date" equals the "Pay to date" (no allowance subtracted). What to do: sign in to the HMRC Personal Tax Account at gov.uk/personal-tax-account, check "Pay and income tax", and either update the employment online or call 0300 200 3300. The refund usually arrives in the next payroll cycle.
This is a modelled study, not a measurement. It is not derived from PayslipIQ user data, HMRC administrative data, or any observed sample of real payslips. We generated 10,000 synthetic UK PAYE payslips — each with a tax code, NI category letter, pension arrangement, and where applicable a student-loan plan — applied HMRC 2026/27 tax tables to them, and then injected errors at assumed rates. The headline 11.4% figure is therefore an output of those assumptions, not an independent discovery: if the assumed rates are wrong, the headline is wrong by roughly the same proportion. We publish the assumptions precisely so readers can judge them.
To be explicit: none of the percentages in Figure 1 (18.2%, 13.7%, 11.4%, 9.8% and so on) is published by HMRC, ONS, CIPP, the Student Loans Company or The Pensions Regulator. They are PayslipIQ modelling assumptions. The cited public sources establish that each failure mode exists and how it operates — CIPP's February 2025 RTI report documents tax-code and data-mismatch failures through employer case studies, HMRC's manuals define how salary sacrifice and NI category letters must work, GOV.UK publishes the student-loan thresholds that make a wrong plan costly — but no public source quantifies the frequency of these errors at payslip level. Where we found no quantitative anchor at all, we chose what we believe is a conservative rate and marked the row accordingly in the dataset. Reasonable people could choose different rates; the model exists to make the consequences of a set of stated assumptions inspectable, not to prove the assumptions.
A 100-row illustrative sample of the synthetic dataset (one row per modelled payslip: salary, tax code used vs. correct, NI category used vs. correct, pension, loan plan, injected error type and annual cost estimate) is published at /data/uk-payslip-errors-2026-dataset.csv. The full 10,000-row dataset, the complete table of injection rates and the generation code are available on request to press@payslipiq.co.uk.
Suggested citation:
PayslipIQ (2026). UK Payslip Errors 2026 — Modelled Anomaly Index (modelled estimate, 10,000 synthetic PAYE payslips, 2026/27 tax tables). Published 17 May 2026. https://payslipiq.co.uk/data/uk-payslip-errors-2026
Link policy: all figures and charts on this page are free to reuse under CC BY 4.0 with attribution to PayslipIQ and a link to this URL. Please describe the headline as a "modelled estimate" — wording such as "a study found 11.4% of UK payslips are wrong" misrepresents the method, and we will ask for a correction if we see it.
All figures on this page are free to cite with attribution to PayslipIQ — UK Payslip Errors 2026 and a link to this URL, provided they are described as modelled estimates (see the suggested citation). The full methodology, the assumption table and the underlying dataset are available on request.
Contact: press@payslipiq.co.uk. See the PayslipIQ press room for spokespeople and brand assets.
All figures are modelled estimates based on published HMRC, ONS, SLC, CIPP and Pensions Regulator data. They are not derived from individual user payslips. PayslipIQ does not store payslip content — see our Privacy notice and Trust Centre.