HolidayProof

1. One-liner

Rebuilds six years of holiday entitlement and pay for irregular-hours staff, and proves the numbers to the Fair Work Agency.

2. Trend signal — why now?

On 6 April 2026 the Employment Rights Act 2025 turned holiday record-keeping into a statutory duty with a criminal offence attached. Until that date, employers in Great Britain had no legal obligation at all to record holidays taken. Now they must keep records “adequate” to show compliance with leave entitlement, holiday pay at the correct rate, and payment in lieu on termination — for six years — and failure is an offence punishable by potentially unlimited fines (Lewis Silkin, Littler).

The enforcer arrived the next day. The Fair Work Agency launched 7 April 2026. Its proposed holiday-pay regime: notices of underpayment at 200% of arrears, capped at £20,000 per worker, minimum £100 per worker, halved if remediated within two weeks — with a six-year look-back covering non-compliance after 18 December 2025 (Baker McKenzie). That is a shift from “an individual worker sues you” to “a state agency audits your whole payroll for six years.” Separately, tribunal time limits double from three to six months on 1 October 2026.

And the underlying arithmetic is a minefield. Since 1 January 2024 irregular-hours and part-year workers accrue at 12.07% of hours worked, while workers with normal hours must still use the 52-week reference period — and rolled-up holiday pay is permitted for the first group only. Employers routinely apply one method to everybody “for administrative convenience,” which is exactly the error pattern that produces six-figure liabilities: a recent tribunal decision, Littlewood v Nuffield Health, awarded roughly £149,000 (Wrigleys).

Population: 1.23 million UK workers on zero-hours contracts in their main job — a record — with accommodation and food services at 32.2% of workforce on no-guaranteed-hours arrangements (Lancaster University Work Foundation, ONS EMP17).

Provenance:

• Signal 1 (demand): New criminal duty to keep “adequate” holiday records from 6 April 2026, six-year retention, unlimited fines; guidance on what “adequate” means is explicitly absent — https://www.lewissilkin.com/insights/2026/03/27/failure-to-keep-adequate-holiday-records-will-become-a-criminal-offence-from-april-2026 — 2026-03-27
• Signal 2 (economic): Fair Work Agency holiday-pay enforcement strategy — 200% of arrears capped £20,000/worker, six-year look-back, penalties waived if arrears repaid before investigation — https://www.bakermckenzie.com/en/insight/publications/2026/07/united-kingdom-fair-work-agency-holiday-pay-enforcement-strategy — 2026-07
• Signal 3 (feasibility): Method divergence since Jan 2024 (12.07% accrual for irregular/part-year vs 52-week reference period for normal-hours; rolled-up permitted only for the first group) makes recomputation from raw timesheet + payslip data a mechanical, automatable job — https://www.acas.org.uk/irregular-hours-and-part-year-workers/calculating-holiday-pay — 2026-08
• Signal 4 (market size): Record 1.23m zero-hours workers; 32.2% of accommodation and food services workforce on no-guaranteed-hours contracts — https://www.lancaster.ac.uk/work-foundation/news/zero-hour-contracts-reach-new-record-high-as-workers-wait-for-new-rights-to-arrive — 2026 Category: Regulatory arbitrage

3. The opportunity

Every HR and leave vendor in the UK sells the forward-looking balance: how many days does Priya have left this year. Nobody sells the backward-looking proof: on 14 March 2025, when Priya was paid for a day off, was the rate right under the method that legally applied to her contract type on that date — and can you produce the working?

That distinction is the whole business. Under a tribunal claim the worker had to notice and sue. Under the FWA the agency comes to you, asks for six years of records, and the penalty is 200% of whatever it finds. Crucially, the FWA will not penalise an employer who has already repaid the arrears before an investigation begins. That converts a compliance product into a self-interested one: find your own exposure, pay it down quietly, keep the penalty at zero.

Incumbents and why they don’t cover it:

Player	Price	What it does	Gap
BrightHR	from ~£16.67/employee/mo (Core HR)	Absence booking, holiday planner	Forward balances; no retrospective recomputation, no method-by-date logic
Breathe HR	£24–£579/mo by headcount band	Leave requests, approvals	Same — a booking system, not an evidence system
Employment Hero	from £4/employee/mo (HR Essentials)	HR + leave	Same
Payroll bureaux / Sage / Moneysoft	varies	Pay holiday when told to	Compute what payroll was instructed, not what law required
Employment solicitors	£250–£450/hr	Bespoke holiday-pay audits	Correct, unaffordable for a 40-person restaurant group

The advisory market already prices this pain at solicitor rates. Nobody has productised it at £200/month. And the sector-specific warning is on record: “the new regime demands integrated HR and payroll systems capable of producing a defensible audit trail. Fragmented records across departments will prove inadequate under FWA scrutiny.”

4. Target market

• Primary customer: Owner-operator or finance lead at a UK SME with 20–250 staff, of whom 30%+ are irregular-hours, part-year or zero-hours — hospitality groups (2–15 sites), care providers, cleaning and security contractors, retail chains, agencies, seasonal leisure. Secondary buyer, and the better channel: payroll bureaux and accounting firms running payroll for 50–500 SME clients.
• Why they buy: Because “adequate” is undefined, criminal, and they know their spreadsheets aren’t it. Practitioner guidance is blunt that existing payroll and HR systems may be insufficient and that employers must retain supplementary evidence — leave emails, rollover communications — that no HR tool currently files against a worker-year.
• Rough TAM reasoning: ~1.23m zero-hours workers concentrated in employers that use them heavily. Hospitality alone: tens of thousands of multi-site operators. If 20,000 UK SMEs have enough irregular-hours exposure to buy, and 3% convert over three years, that’s 600 accounts. At £2,400 ACV that’s £1.44M ARR — and that ignores the bureau channel, which resells to hundreds of clients per contract.
• Why now for them: The clock started 6 April 2026. Every month that passes adds a month of records they cannot retroactively create, and the FWA’s powers switch on in 2027 with a look-back that already covers the period they’re living through now.

5. Product sketch (MVP)

• Import once: pull timesheets/rotas and payslip history from the payroll or T&A system (Sage, BrightPay, Moneysoft, Xero Payroll, Deputy, Planday) or accept a CSV/PDF payslip dump.
• Worker classification pass: for each worker, for each leave year, decide whether they were genuinely irregular-hours/part-year (12.07% accrual, rolled-up pay permitted) or normal-hours (52-week reference period, rolled-up not permitted) — and show the evidence behind the call, because this is where employers get it wrong.
• Recompute and diff: rebuild what should have accrued and what should have been paid, week by week, against what was actually paid. Output: a per-worker arrears figure and a total exposure number with a confidence band.
• Remediation pack: a per-worker breakdown suitable for a corrective payment run, plus the wording — since arrears repaid before an FWA investigation carry no penalty.
• The defence file: an exportable, dated, six-year-retention record set — entitlement, leave taken, pay rate applied, method used and why, termination payments — the artefact “adequate records” actually means.
• Drift watch: ongoing monthly re-run flagging workers whose pattern has changed enough to move them between methods, or whose rolled-up pay is no longer lawful.
• Evidence attachments: file leave request emails, rota screenshots and rollover approvals against the correct worker-year so the trail isn’t fragmented across departments.

6. AI angle — what’s load-bearing

Two places, both structural.

One: the classification judgement. “Irregular hours or part-year worker” is not a database field — it’s a legal characterisation of a work pattern, and the failure mode on record is an employer paying rolled-up holiday to someone who “worked standard hours every week.” Deciding it means reading a contract, looking at 52 weeks of actual hours, and reasoning about consistency. An LLM reading contract PDFs plus the hours series, producing a call with a written rationale a solicitor can check, is doing the work an employment lawyer bills £350/hour for.

Two: ingesting the mess. These employers don’t have clean data. They have PDF payslips, a rota in Excel per site, a WhatsApp thread where a shift got swapped, and a leave request in an inbox. Extracting a normalised hours-and-pay series from that heap is exactly what vision-capable models became good enough at in the last 18 months.

Strip the AI out and you have a calculator that requires perfectly clean structured input — which is precisely what the customer does not have. That’s why the incumbents’ “holiday calculators” don’t solve it.

7. Localisation angle

This is the localisation play — UK-only, and deliberately so. The 12.07%/52-week split, the Harpur Trust aftermath, rolled-up holiday’s re-legalisation for one worker class only, the FWA’s penalty schedule: none of it ports. A US or EU generalist HR vendor will not build this depth for one country’s SME band, which is the moat’s whole basis. Natural adjacency later is Ireland (different rules, same shape) — not a v1 concern.

8. Business model — path to $1M–$5M ARR

• Pricing: £149/mo (up to 50 workers), £299/mo (51–150), £599/mo (151–400). Plus a one-off historic reconstruction fee of £750–£2,500 depending on headcount and data mess — this is the thing they actually want first, and it prices against the solicitor hour.
• Bureau tier: £1,200–£4,000/mo for payroll bureaux, priced per client-book band, white-labellable into their client portal.
• ACV: ~£2,400 direct SME; ~£24,000 bureau.
• £1M ARR math: 250 direct SMEs at £2,400 (£600K) + 17 bureaux at £24,000 (£408K) = ~£1.0M. Reconstruction fees sit on top as non-recurring cash that funds the sales motion.
• £5M ARR math: needs the bureau channel to carry it — ~120 bureaux at £24K (£2.9M) plus 850 direct SMEs (£2.0M). Plausible only if the FWA actually starts issuing notices in 2027 and the first penalty makes trade press.
• Expansion path: headcount bands, additional sites, then adjacent record duties under the same Act (the FWA’s remit spans NMW, statutory sick pay and agency worker rules — same data, same evidence shape).

9. Go-to-market wedge — first 100 customers

1. The exposure teardown, sent cold. Pick UK hospitality and care groups with 40–200 staff from Companies House filings cross-referenced against Food Standards Agency site registers (multi-site = irregular-hours exposure). Send a two-page “what your holiday-pay exposure probably looks like” using published sector accrual patterns, offering a free reconstruction of one worker-year from a single payslip PDF. That single-worker free run is the demo. Target 1,500 sends, 6% reply, 100 free runs, 20% close.
2. Payroll bureaux, one at a time. There are hundreds of independent UK payroll bureaux whose clients are exactly this population and who are getting the “what does adequate mean?” question right now with no answer. Offer them the tool free for their own book for 60 days in exchange for a case study. One bureau contract = dozens of end-clients.
3. Employment solicitors as referrers. Firms doing holiday-pay audits at £350/hour hate the data-cleaning half of the job and can’t bill it well. Give them the reconstruction engine free; they keep the legal opinion and the client. Ten referring firms is a pipeline.
4. Ride the first FWA notice. The moment the FWA issues its first published notice of underpayment for holiday pay, that’s the content event — a same-week teardown of the case posted to Personnel Today, AccountingWEB and the CIPP community, with a free self-assessment tool attached.
5. The trade bodies. UKHospitality, the Care Provider Alliance and the Recruitment & Employment Confederation all run member webinars on ERA 2025 compliance. A free “are your holiday records adequate?” audit for members is a standard slot to buy into.

10. Build complexity — justification

Medium. The recomputation engine is deterministic arithmetic with a nasty rules matrix (which method, which date, which worker class, carry-over, rolled-up legality, termination PILON) — a few weeks of careful work plus a test suite built from Acas worked examples. The hard part is ingestion: payroll integrations and PDF payslip extraction across a dozen formats. Off-the-shelf everywhere — vision models for extraction, standard web stack, no infra research. Two people, 14–18 weeks to a v1 that can run a real 60-person restaurant group end to end. The classification layer needs a real employment-law reviewer on retainer, not a model alone.

11. Gating checklist

Gate	Pass?	Note
Legal in target market	✅	Compliance tooling; not regulated advice if opinions stay attributable to the employer’s own adviser
Ethical — no harm / dark patterns	✅	Product’s output is workers getting arrears they’re owed
Market exists (evidence above)	✅	Criminal duty live since Apr 2026; enforcer live; incumbents charge £350/hr for the manual version
1–5 person team can build this	✅	Two builders + fractional employment-law reviewer
Launchable with <$50K / ₹40L	✅	Main costs are the legal reviewer and outbound

12. Feasibility score

Axis	Weight	Score	Notes
Problem intensity	20	17/20	Criminal offence, unlimited fine, 200%-of-arrears penalty, six-year look-back. Not hair-on-fire today only because FWA powers land in 2027 — but the record duty is already live and unrecoverable if skipped.
Demand evidence	15	13/15	Multiple independent legal-practice alerts, a live enforcement consultation with a published penalty schedule, existing paid advisory at solicitor rates, £149K tribunal award. Missing: direct verbatim buyer complaints — the population is too busy to post.
Build feasibility	15	11/15	Rules matrix is tractable; ingestion across payroll formats is the real work. 14–18 weeks for two.
Distribution clarity	15	13/15	Named channels: bureaux, employment solicitors, Companies House + FSA site lists, three trade bodies. Free single-worker reconstruction is a cheap, concrete demo.
Revenue mechanics	15	12/15	Pricing benchmarks cleanly below solicitor hours and above HR-tool per-seat norms. One-off reconstruction fee funds CAC. Bureau tier is the £5M lever and is unproven.
Time to first revenue	10	8/10	The historic reconstruction is sellable before the SaaS exists — a paid manual audit in month one is legitimate.
Defensibility	10	4/10	Execution and depth-of-rules moat only. BrightHR could ship a version. The bet is that they won’t build retrospective liability tooling that implicates their own existing customers’ past data.
Total	100	78/100

13. Qualitative modifiers

Founder-fit tags

technical-heavy · sales-heavy · domain-expertise-required

Key assumptions to validate

1. Assumption: SMEs will pay to discover their own arrears rather than stay wilfully ignorant. How to test: offer 30 free single-worker reconstructions; measure how many ask for the full-book run once shown a non-zero figure. If <25% proceed, the whole thesis is wrong.
2. Assumption: payroll bureaux want to be the channel rather than viewing exposure discovery as a liability to their own past work. How to test: 15 structured calls with independent bureaux; ask directly whether they’d surface historic errors to clients.
3. Assumption: the raw data is extractable at acceptable accuracy from real customer mess (PDF payslips + Excel rotas). How to test: run 10 real anonymised data dumps end to end; require ≥95% hours-series accuracy against manual reconstruction.
4. Assumption: FWA enforcement actually arrives in 2027 rather than slipping. How to test: track the FWA consultation response and commencement regulations quarterly; a two-year slip guts urgency.

Risk flags

1. Enforcement timing: the FWA’s holiday-pay powers are not yet in force and no commencement date is set. If they slip to 2028, urgency collapses and this becomes a nice-to-have.
2. Self-incrimination reflex: a product whose first output is “you owe your staff £84,000” can be refused for exactly that reason. The remediation framing — penalties waived if repaid before investigation — has to be the lead message, not a footnote.
3. Incumbent absorption: BrightHR, Employment Hero and the payroll vendors will eventually add a “holiday records compliance” tab. The defence is retrospective depth and the evidence artefact, not the balance tracking.
4. Advice boundary: classifying a worker’s contract type shades toward legal advice. Structure output as findings the employer’s own adviser signs, and keep an employment lawyer in the loop.

14. Structured verdict

Score:                  78/100
Verdict:                GO
Confidence:             Medium
Best-fit builder:       Technical founder + a UK employment-law or payroll operator; needs someone who
                        can sell to accountants and hospitality finance leads, not just build
Time to revenue:        4–8 weeks (paid manual reconstruction audits before the product ships)
Capital to launch:      £25,000–£35,000 (legal reviewer retainer + outbound)
Top 3 assumptions to validate first:
  1. SMEs proceed to a paid full-book run after a free single-worker reconstruction shows non-zero arrears (30 free runs, need >25% conversion)
  2. Payroll bureaux will channel rather than block (15 structured calls)
  3. Real-world data is extractable at ≥95% accuracy (10 anonymised end-to-end dumps)
Kill criteria:
  - Abandon if fewer than 25% of free-reconstruction recipients with non-zero arrears request the paid full run
  - Abandon if FWA holiday-pay enforcement powers are formally deferred beyond 2028
  - Abandon if two of the three top HR vendors ship retrospective liability recomputation before month 9

15. Next step — 1-week validation sprint

• Day 1–2: Build the rules matrix by hand in a spreadsheet and validate it against every Acas worked example for irregular-hours and 52-week-reference cases. If the matrix can’t reproduce Acas’s own answers, stop.
• Day 3: Source 5 real anonymised payslip-and-rota dumps from a friendly hospitality operator and a friendly bureau. Reconstruct one worker-year manually for each. Time yourself — if it takes more than 90 minutes per worker manually, the automation value is large and the pricing holds.
• Day 4: Call 10 payroll bureaux cold. Single question: “When a client asks what ‘adequate holiday records’ means, what do you tell them?” Count how many have an answer.
• Day 5: Send the free single-worker reconstruction offer to 40 multi-site hospitality and care operators pulled from Companies House.
• Go/no-go: proceed only if ≥8 of 40 accept the free reconstruction and ≥4 of 10 bureaux say they have no answer for their clients. Anything less means the pain isn’t felt yet — revisit when the FWA issues its first notice.