Nexus Track Direct.
Implementation

The ERP Go/No-Go Checklist: 30 Checks to Run in the Final Two Weeks Before Cutover

Most troubled go-lives were approved by a steering committee that never wrote down what "ready" meant. Here are 30 checks, each with a pass threshold, an owner and a written rollback trigger.

Nexus Track Direct editorial team · November 2026 · 10 min read

Who makes the go/no-go call, when, and on what evidence

Two weeks out, your steering committee has to answer one question: do you switch on the new system on the planned date or not? Most teams answer it with status reports colored green, amber and red. That's the wrong evidence. An ERP go-live checklist only helps if every line has a numeric pass threshold, one named owner and evidence someone can inspect, such as a signed reconciliation, a defect export or a bank acknowledgment file.

Hold three checkpoints. At T-14, run a preliminary review to find what's at risk. At T-7, run the formal readiness assessment against every check below. At T-1, after the legacy freeze and before data loads begin, take the final vote. The executive sponsor, usually the CFO or COO, owns the decision. Process owners and the cutover manager give recommendations. The systems integrator advises but doesn't vote, because it has a commercial interest in hitting the date.

Advertisement

Each of the 30 checks is marked Gating or Conditional. If any gating check fails, the answer is no-go. A conditional check can fail only if there's a documented workaround, an owner and a fix date. Agree on this classification at T-14, not in the final meeting. Criteria renegotiated at T-1 almost always loosen, and that's how weak go-lives get approved.

Reconciled balances: checks 1 to 7

Finance data is where a bad cutover does the most lasting damage. A wrong opening balance corrupts every report built on it, and auditors tend to find the problem months later. Run these checks on the data from the final mock cutover, then run them again during the live load.

#CheckPass thresholdOwnerType
1Trial balance by legal entity, legacy vs. new after account mappingZero variance at account level. The only tolerance is documented roundingGroup controllerGating
2Open receivables tie to control account and aged listing100% of value and item count, including credits and unapplied cashAR leadGating
3Open payables tie, including held invoices and prepayments100% of value and countAP leadGating
4Inventory quantity by site and bin, plus valuation100% quantity match on counted items. Value variance under 0.1%, every difference explainedInventory controllerGating
5Fixed asset register and one-month test depreciation runCost, accumulated depreciation and net book value tie by asset classFixed asset accountantConditional
6Open purchase and sales orders, including partially received or shipped lines100% of count and open valueProcurement and order management leadsGating
7Foreign currency revaluation and intercompany balancesRevaluation matches legacy at the same rate. Intercompany nets to zeroTreasury leadConditional

Inventory is the check most likely to fail. Schedule a physical or cycle count close to the freeze so you're reconciling against stock that's really on the shelf, not just against legacy records.

Open defect severity and test exit: checks 8 to 12

Agree on severity definitions in writing before anyone counts defects. A workable standard: Severity 1 blocks a business process and has no workaround. Severity 2 blocks a process, but a tested manual workaround exists. Severity 3 is an inconvenience.

  1. Severity 1 defects (Gating, test manager): zero open. No exceptions, including "fix due in week one."
  2. Severity 2 defects (Gating, test manager): no more than an agreed cap, often in the range of 5 to 10 for a mid-sized rollout. Each needs a documented workaround that affected users have been trained on.
  3. End-to-end scenario sign-off (Gating, process owners): every critical scenario passes, covering order-to-cash, procure-to-pay, record-to-report and plan-to-produce, plus 95% or more of all scenarios. Business process owners sign, not the consultants.
  4. Final-build regression (Gating, release manager): the last transport or deployment introduced no new Severity 1 or 2 defects, and the code freeze is in force.
  5. Performance at production volume (Conditional, technical lead): MRP, warehouse wave release, invoice runs and period-end jobs all finish inside agreed windows. MRP, for example, should finish before the morning shift starts.

Watch for Severity 1 defects downgraded in the final week with no new workaround. Ask what changed in the software, not what changed in the label.

Cutover rehearsals and the downtime window: checks 13 to 17

Your cutover plan is only as believable as its last rehearsal. A team that has rehearsed on a data subset doesn't know how long the real load takes.

#CheckPass thresholdOwnerType
13Full mock cutovers completedAt least two, with the final one on production-sized data using the live runbookCutover managerGating
14Rehearsed duration vs. downtime windowRehearsal takes 75% or less of the window. Example: a 40-hour weekend window allows no more than a 30-hour rehearsalCutover managerGating
15Runbook completenessEvery task has an owner, a predecessor, a duration and an out-of-hours contact. No task owned by "TBD"Cutover managerConditional
16Data load rejection rate in final mockUnder 0.5% of records rejected. All rejects categorized, with scripted fixesData migration leadConditional
17Freeze and blackout communicatedLegacy freeze time, last shipment and last invoice run are confirmed with customers, suppliers, carriers and sitesOperations leadGating

Choose the date with care. Go live at the start of a period, but not at fiscal year-end and not in peak season. Fiscal years differ by country: April to March is common in India, July to June in Australia, and calendar year across much of Europe and Latin America. A multinational may need different dates for different entities.

Bank and tax-authority integrations: checks 18 to 23

External integrations need their own scrutiny because you control your system but not the bank or the tax authority. Their onboarding queues, certificates and test environments run on their timelines, and the rules vary by country. Run these checks for each jurisdiction.

  • 18. Payment files (Gating, treasury): each bank has accepted a test file in its test environment for every format you use, such as ISO 20022 pain.001, SEPA, BACS in the UK, NACHA in the US, or NEFT/RTGS in India. Where the bank allows it, send one low-value live payment per account and format.
  • 19. Statement import (Conditional, treasury): camt.053, MT940 or BAI2 statements load, and auto-matching hits an agreed rate on a sample week. Many teams aim for roughly 80%, but set your own target.
  • 20. E-invoicing and clearance (Gating where mandatory, tax lead): production credentials, digital certificates and provider onboarding are complete. Examples include the IRP and e-way bills in India, ZATCA integration in Saudi Arabia, CFDI through a PAC in Mexico, NF-e in Brazil, SdI in Italy and KSeF in Poland. Mandates and deadlines change, so confirm current requirements locally. In clearance regimes, an invoice that can't be cleared can block shipment of goods.
  • 21. Tax determination (Gating, tax lead): a test matrix covering VAT, GST or sales tax, reverse charge, exemptions and withholding gives correct results in 100% of cases. Have a qualified tax adviser in each relevant country review the matrix.
  • 22. EDI, carrier, 3PL and WMS links (Gating for top partners, integration lead): the trading partners that carry the bulk of your order volume, often about 80%, have tested orders, ASNs, labels and invoices end to end.
  • 23. Statutory and audit readiness (Conditional, controller): external auditors have agreed how opening balances will be signed off, and legacy data is archived read-only for the retention period each country requires. Periods commonly run from several years to a decade or more, so check with local advisers.

User and operational readiness: checks 24 to 27

Software that works but that people can't use correctly still fails. It just fails more slowly, through miskeyed orders and spreadsheet workarounds.

  • 24. Training (Gating, change lead): 90% or more of named end users trained and passed a role-based assessment. 100% of super users.
  • 25. Access and segregation of duties (Gating, IT security): all production roles are provisioned and every user has logged in to test their access. SoD conflicts have been removed or formally accepted. No shared generic accounts.
  • 26. Hypercare staffing (Conditional, program manager): named super users on every site and every shift for the first two to four weeks, plus a triage path with response times.
  • 27. Business continuity buffer (Conditional, operations and finance): safety stock built or shipments pulled forward, key customers notified, and critical supplier payments and payroll run before the freeze.

The rollback plan and its written trigger: checks 28 to 30

Many programs have an ERP rollback plan on paper. Few have written down who can call a rollback and on what evidence. Without that, nobody wants to make the call at 3 a.m., and the team presses on out of hope.

  • 28. Rehearsed rollback and point of no return (Gating, cutover manager): the legacy system can be restored to its frozen state, and the restore has been tested. The runbook marks the point of no return, usually the first irreversible external event: a bank file transmitted, an e-invoice cleared or a shipment confirmed to a customer. A cleared e-invoice generally can't be deleted, only reversed with a credit note.
  • 29. Written trigger with a named decision-maker (Gating, executive sponsor): state the criteria in numbers. For example: "If the trial balance doesn't reconcile by hour 24, or any Severity 1 defect in order-to-cash is still open four hours before the point of no return, we roll back. The CFO decides, and the cutover manager is the alternate."
  • 30. Fix-forward plan after the point of no return (Conditional, program manager): agreed manual fallbacks, such as manual bank payments, invoicing outside the system where local rules permit it, and paper picking, each with an owner and a maximum duration.

Scoring the vote and pricing a delay

Add up the results and apply the criteria you agreed at T-14. Don't reopen them in the meeting.

OutcomeConditionWhat follows
GoAll gating checks pass, no more than 3 conditional fails, each with owner and dateProceed with the cutover plan as written
Go with conditionsAll gating checks pass, 4 to 6 conditional failsSponsor signs a written risk acceptance. Hypercare staffing goes up
No-goAny gating check fails, or more than 6 conditional failsSet a new date tied to specific fixes, not to a calendar slot

Delay costs real money, so put a figure on it. Here's a hypothetical example: a 300-person distributor with an eight-person integrator team at a blended $120 to $180 an hour is spending roughly $40,000 to $60,000 a week. If the next clean period start is four to five weeks away, add legacy support running in parallel, and a slip could land somewhere around $200,000 to $350,000. These are indicative 2026 ranges that vary widely by contract and region, so check your own statement of work. Set that against two weeks of not being able to invoice or ship, and the delay usually looks cheap.

The decision rule: if any gating check fails, you slip, whatever the delay costs. This week, give every one of the 30 lines an owner, put the T-14, T-7 and T-1 checkpoints in the sponsor's calendar, and have the gating and conditional classification signed before anyone sees the first result.

Frequently asked questions

Who should make the final ERP go/no-go decision?

The executive sponsor, usually the CFO or COO, should own the decision. Process owners, the cutover manager and IT give recommendations. The systems integrator should advise but not vote, because it has a commercial interest in keeping to the date.

How far before cutover should the go/no-go meeting happen?

Use three checkpoints: a preliminary review at T-14, a formal readiness assessment at T-7, and a final vote at T-1 after the legacy freeze and before data loads begin. If you only hold the final meeting, you won't have time to fix anything it uncovers.

How many open defects are acceptable at go-live?

Zero Severity 1 defects, meaning nothing that blocks a process without a workaround. Most programs allow a small, agreed cap on Severity 2 defects, often 5 to 10, as long as each has a tested workaround that users have been trained on. Severity 3 items can usually wait for hypercare.

Can we roll back after go-live?

Yes, but only before the point of no return. That's usually the first irreversible external event, such as a bank file being sent, an e-invoice being cleared by a tax authority, or a shipment being confirmed. After that, you'll generally have to fix forward using manual workarounds, so write that plan down in advance.

Is year-end a good time to go live on a new ERP?

Usually not. Go live at the start of a period, but avoid fiscal year-end and peak trading season, when close and audit workloads are heaviest. Fiscal years vary by country, so multinationals may need different dates for different entities.

How much does delaying an ERP go-live typically cost?

Most of the cost is integrator fees plus running legacy and new systems in parallel. In the hypothetical mid-sized example above, that's roughly $40,000 to $60,000 a week and $200,000 to $350,000 for a one-period slip. These are indicative 2026 ranges that vary by contract and region, so confirm against your own statement of work.

Written by the Nexus Track Direct editorial team. Prices, features and rules change often - confirm details with the vendor or a licensed professional before you decide.