Go/No-Go at Cutover: Four-Country Payroll Implementation with Parallel Run Defects Open
Meridian Consumer Goods (4,200 employees across the UK, Ireland, the Netherlands and Poland) is replacing four local bureau arrangements with a single global payroll platform run by a new managed service provider. Go-live is fixed for 1 April because the UK tax year starts on 6 April and the legacy licence expires on 30 June; the incumbent bureau must receive four weeks' notice and has already declined to extend. The programme is in parallel run two of three. Open defects: 11 UK employees carry the wrong NI category letter, producing a £4,300 net variance; 38 Irish employees have no PRSI class change applied after a mid-year move from part-time hours; six Dutch assignees have no 30% ruling applied, a €9,100 net shortfall; and PPK contributions were not calculated for 212 Polish rehires. Nineteen defects are open overall, seven untriaged. The steering group meets in five working days and the sponsor has asked for a go/no-go recommendation with supporting evidence.
1. Formulate the go/no-go recommendation and defend the decision logic you would put to the steering group.
Read the worked answer
The decision is not pass/fail but risk allocation, and three of the four options are constrained: full delay is blocked by the legacy licence and the incumbent's refusal to extend, so the real choice is full go-live, scope-reduced go-live, or go-live into shadow operation for April.
Start from what parallel run exists to prove. Its purpose is to demonstrate that the control environment detects errors before employees are paid, not that zero errors exist. Every named defect was found by reconciliation, which is positive evidence. What must be tested is the trend: are detected defects falling between cycle one and cycle two, and are they configuration (fixable once) or data (population-wide and recurring)?
Apply a written materiality threshold, agreed before the meeting: any defect that affects a statutory filing or rate, any defect affecting more than 1% of a country population, and any net variance above an agreed euro amount per employee must be closed and retested. On that basis the NI category letters and PRSI classes are statutory, so they close or the country does not go live. The 30% ruling is employee-visible net pay and must close. PPK on 212 rehires looks population-wide, so the credible option is deferring Poland four to six weeks onto a paid legacy extension.
Recommendation: conditional go-live for UK, Ireland and the Netherlands, Poland deferred, contingent on a clean retest of the four named defects, seven untriaged defects triaged within 48 hours, a paid transition-services extension with the incumbent to 31 July, and one April cycle run in shadow on the new platform.
2. Design the control gates the final parallel cycle must clear, and the evidence you would require at each gate.
Read the worked answer
Gates should be few, binary and evidenced, because a weak gate is worse than none.
Gate 1 - Data migration integrity. Record counts and control totals for headcount, YTD taxable pay and tax, absence balances, garnishments and bank details, reconciled source-to-target with named sign-off. For the UK this is decisive: the 6 April start means opening YTD balances and tax codes carry the employee's full-year position, and a migrated code error surfaces at year end, not at go-live.
Gate 2 - Full-population gross-to-net comparison. Every employee, every country, at line level, not sampling. Tolerance of zero on statutory deductions and an agreed tolerance on net. Variances above tolerance must be explained by an identified defect or by a documented reason (new starter, leaver, mid-month change), not by 'expected rounding'.
Gate 3 - Statutory output testing. RTI submission files, PRSI and PPK contribution files, Dutch aangifte, and payment files tested end-to-end into the receiving authority or bank format, including one zero-value prenote to prove bank connectivity.
Gate 4 - Defect closure evidence. Each defect closed with a retest script and a named tester independent of the person who fixed it, plus an exported defect log with severity, root cause category and trend.
Gate 5 - Operational readiness. Named country owners, hypercare roster for two cycles, documented manual workarounds, and defined rollback criteria. Acceptance should be defined contractually, so that go-live does not constitute deemed acceptance of the platform.
3. What commercial and contractual levers would you use with the incumbent bureau and with the new provider, and why?
Read the worked answer
Leverage is asymmetric and time-limited, so sequence matters.
With the incumbent: your leverage is highest now, before they know the notice is certain. Buy a transition-services extension to 31 July at a fixed price before disclosing the deferral decision, and secure exit assistance - final data extracts, YTD files, scheme leaver notifications and a named contact. Without an exit-assistance clause the incumbent can hold the migration hostage in June at whatever price it likes. Also confirm what happens to the 30 June licence expiry if the extension is not signed; that is the real drop-dead date, not 1 April.
With the new provider: the contract should carry (i) country-by-country acceptance criteria tied to the gates above, so acceptance is a defined test rather than elapsed time; (ii) service credits and an at-risk fee percentage - commonly 10% - retained until two clean live cycles; (iii) defect severity definitions with fix SLAs and a warranty period covering configuration errors discovered after go-live; (iv) an obligation to support parallel or shadow operation in April at no extra charge; (v) audit and data-return rights, given payroll data is personal data and the provider will hold it.
Finally, take the Poland deferral as a formal change request with a priced scope, not a verbal concession - an unpriced scope change becomes an argument in month three.