Testing and go live

ERP testing and go live assurance

ERP programmes rarely fail because a transaction would not post. They fail because testing proved the software and nobody proved the business. We plan ERP testing around your processes and your data, rehearse cutover until it is dull, and give the board an honest go or no go.

19 yrssenior systems delivery
7finance platforms delivered
39 daysearly on an Oracle to Dynamics migration
22m+assets in the National Grid separation

Delivered for and alongside

ShellBPCapitaBaringaOfgemSmart DCCInsightsFlui TechnologiesExxonMobilVodafoneConstellation SoftwareDeloittePwCEYNational GridElexon

A test strategy that proves the business, not the software

A test strategy is a set of decisions taken before build finishes, not a document produced when the plan says testing starts. The decisions: which test levels you will run, who owns each, what data they use, what counts as an entry and an exit criterion, and how defects are classified. SAP's method distinguishes unit tests of a single piece of a process, string tests that chain units together, full business process tests run from standard scripts, formal integration and data migration tests, user acceptance testing run by the customer's own line of business experts, and regression testing after each release. Microsoft's go live checklist asks for signed off system integration, user acceptance and performance testing before anyone discusses a date. The vocabulary differs by vendor; the logic does not.

Three choices separate test strategies that work from those that merely exist.

  • Test business scenarios, not configuration. A scenario is an order with a credit hold, a part delivery, a return and a credit note; it crosses modules and interfaces. Module by module testing passes while the quarter end fails.
  • Use migrated data. Test cases run on clean demonstration data prove nothing about your real supplier master, duplicates and all. From integration testing onwards, every cycle runs on a fresh mock migration.
  • Include roles. Security and role testing confirms that each user can reach what they should and nothing more, and it finds the approver who cannot approve before the first live purchase order does.

We write the strategy with you and your systems integrator, then hold both sides to it. The ERP implementation plan article shows where testing sits in the wider plan, and why squeezing it is the most common response to a build that is running late.

Reconciliation and integration testing

Data migration reconciliation is the part of ERP testing most often done badly, because it is tedious and because it exposes decisions nobody wanted to take. We reconcile at three levels. Counts: records extracted, transformed, loaded and rejected, by object, with every rejection explained. Control totals: general ledger balances by company code and account, open receivables and payables by customer and supplier, stock quantities and values by location, fixed asset cost and net book value, all agreed to the legacy trial balance on the same date. Samples: a sensible set of records checked field by field by the people who own the data, with the sign off recorded. A mock migration that reconciles to the penny but takes a week to load is still a problem, so timing is measured on every cycle.

Interfaces fail at go live more often than the core system, because they are tested last and tested lightly. Integration testing has to cover each interface at realistic volumes and with deliberate failures: a bank file rejected, a web order with a missing postcode, a payroll feed posted twice. We want to see what the system does with each, who is alerted and how the transaction is replayed. Microsoft's guidance on performance testing applies here as well: form load times, search and integration performance under a realistic production load, not one user on a quiet environment.

Where the programme has an integration platform or a tangle of point to point links, our ERP integration service covers the design, and our data migration consultancy covers extraction, cleansing and reconciliation itself. On this page the point is narrower: neither is finished until the evidence is in the test record, and the test record is what the go or no go meeting reads.

Cutover rehearsals, defect triage and the go or no go decision

Cutover is a short, hostile window. Microsoft puts it at usually less than 48 hours, during which neither the old system nor the new one can be used, and advises against cutting over when you have to close your books or report results. The cutover plan lists every task in order with its timing, an owner and a backup owner, the instructions and tools, a verification and sign off step, and a rollback plan. Then you rehearse it: a mock cutover as close as possible to the real one, with the same data, tools, people, steps and timing, repeated until the team is confident. Each rehearsal fixes the plan, trims the window and shows who is really needed at three in the morning.

Defect triage runs daily through the test phases against a severity scale agreed at the start: what stops go live, what needs a workaround, what can wait for the first release afterwards. The temptation in the final weeks is to downgrade severities to protect the date. We keep the scale honest and publish the list.

The go or no go decision is taken by named people against criteria agreed months earlier, not at a late night meeting under pressure from a calendar. Our criteria usually include: no open defects above the agreed threshold without an approved workaround; integration, acceptance and performance testing signed off; the final mock migration reconciled; training complete and security roles assigned; licences available and assigned; a production support plan signed; and the cutover plan rehearsed. Microsoft FastTrack runs its go live readiness review as a workshop with all parties, and we recommend the same format whatever the vendor, with the sponsor in the room. An ERP go live postponed on evidence costs weeks. One forced through costs the year. When a programme is already past that point, our ERP programme recovery service picks it up.

ERP hypercare that ends on evidence

ERP hypercare is the short period after go live when extra resource and attention are given to users and processes, and the word short matters. It ends on exit criteria agreed in advance, not when the budget runs out or the partner moves to the next client. Microsoft's guidance suggests criteria such as no critical issues open, key processes running efficiently, service levels met and the support team resolving most issues without extra help. We add the first period end close, because month end is where a finance system is truly tested.

Getting there takes preparation that starts in testing. The support team practises the real support process during user acceptance testing, with the project team acting as the resolving authority, so that handover is gradual rather than an event. The contract with your implementation partner should name the people retained through hypercare and the terms on which they stay, because otherwise they will be busy on the next phase. Documentation, task guides and the known issues list are handed over before go live, not promised after it.

We work in two ways: as independent assurance of your partner's testing and cutover, or as the delivery team that runs it from the client side. On the BP Pulse programme, where PEF held the main contract, the Oracle to Microsoft Dynamics migration completed 39 days ahead of forecast, which is the kind of result rehearsal and reconciliation discipline exist to produce. Training and adoption, which decide whether hypercare is quiet or chaotic, are covered on our ERP change management page. Our fees are shown without ambiguity on the pricing page.

What we have delivered

  • Ran the procurement of every software provider on the National Grid separation, more than 22 million assets, through to implementation including Oracle to SAP S/4HANA.
  • Delivered the Oracle to Microsoft Dynamics migration at BP Pulse 39 days ahead of forecast, with a £25m saving, the largest on a programme where PEF held the main contract.
  • Delivered the £9bn Energy Bills Support Scheme for Ofgem, BEIS and HMRC.
  • Implemented and delivered most major finance systems over 19 years: Oracle, Microsoft Dynamics, NetSuite, Sage, SAP S/4HANA, Unit4 and Access.
  • A suite of consultants who have led and analysed pre, mid and post migration decisions and implementations across ERP and CRM platforms, independent of every vendor.

Questions we are asked

What does ERP testing include?

Unit tests of individual configuration and process steps, string tests that chain them, integration or end to end tests across modules and interfaces, data migration tests with reconciliation, user acceptance testing run by the business, performance testing under realistic load, security and role testing, and regression testing after each vendor release. The order and names vary by vendor, but a credible programme runs all of them on migrated data with agreed entry and exit criteria.

What should an ERP go live checklist contain?

Signed off integration, user acceptance and performance testing; a reconciled final data migration; a cutover plan with owners, timings, verification steps and a rollback plan, rehearsed at least once; training completed and security roles assigned; licences available and assigned; a production support plan with escalation and out of hours cover; open defects below the agreed threshold with approved workarounds; and a named group who make the go or no go decision against those criteria.

How do you decide go or no go for an ERP go live?

Against criteria agreed months before the date, by named people, in a readiness review with every party present. The criteria cover defect thresholds by severity, test sign offs, data reconciliation, training and role assignment, support readiness and a rehearsed cutover. The decision should be made on evidence in the test and migration records, not on a status report. Postponing on evidence costs weeks; forcing a go live through on hope can cost a year.

What is ERP hypercare and how long should it last?

Hypercare is the short period after go live when extra resource and attention support users and business processes while early issues are resolved. It should end on exit criteria set in advance, such as no critical issues open, key processes running, service levels met and the support team resolving most issues unaided. For a finance system we recommend running it through the first period end close, since month end is where the system is properly tested.

How many mock cutovers do we need before an ERP go live?

Enough that the last one is dull. Vendor guidance says to rehearse several times until the team is confident, using the same data, tools, people, steps and timing as the real event. In practice we expect at least two full rehearsals, the final one on a complete fresh copy of production data with the cutover window timed, and the plan corrected after each. A rehearsal that overruns is a result, not a failure; it is why you rehearse.

Is your go live date a plan or a hope?

Share your test plan and cutover approach and we will tell you, independently, whether the evidence supports the date.

Book a conversation