Nine weeks from a fixed go-live on a programme that has consumed 84% of budget
Fabrikam Utilities is nine months into an eleven-month Dynamics 365 Finance implementation. The systems integrator reports 58% of 214 backlog items complete, but 17 remain marked critical, including licence-plate receiving on the warehouse mobile app and a three-way match extension for freight accruals. There are 412 open defects, 64 of them severity A. Two of four planned test cycles have slipped. The Tier-2 environment has not been refreshed from production since month three, so testers are validating against stale master data and have stopped logging some defects as "known data issues". The programme has consumed 84% of its implementation budget. Go-live is fixed: the legacy system's hosting contract ends on 31 March and the regulator has been notified of the cutover. The CIO wants a written recommendation in two weeks on whether to proceed.
1. What do you recommend regarding the go-live date, and how do you get to a defensible answer in two weeks?
Read the worked answer
Refuse to answer yes or no until there are written exit criteria. In the two weeks, produce three things.
First, a blocker list: take the 64 severity A defects and classify each as data, configuration, or code, and whether it blocks a business process or a control. Typically fewer than a third survive that filter, and the 17 "critical" backlog items usually contain three that are genuinely launch-critical. Second, a quantified trade-off: for each surviving blocker, state the cost of fixing before 31 March against the cost of a workaround or manual control for 90 days. Third, environment remediation: refresh Tier-2 from masked production data immediately, because every result produced so far is of unknown validity and will not support a go/no-go.
Then present options rather than a verdict: (a) full scope on 31 March, only if the blocker list is empty and one clean end-to-end cycle passes; (b) phased cutover, moving freight accruals and licence-plate receiving to a second wave with a named manual control in the interim; (c) delay with the legacy contract extended, priced and approved, and the regulator re-notified.
Also give the CIO the diagnostic that matters: 84% of budget spent with 42% of scope outstanding is a scope-control problem, not a testing problem. Any recovery plan that does not change scope merely moves the overrun past go-live.
2. How should the architect handle environment and ALM hygiene, and how should the relationship with the systems integrator be reset?
Read the worked answer
Two workstreams that must run together.
Environment and ALM. Establish a build-to-Tier-2 pipeline so every fix is deployed the way it will reach production; manual deployment through the UI cannot be evidenced or repeated, and that is likely today's failure mode. Tier-1 for developer unit tests with ATL/SysTest in the pipeline; Tier-2 rebuilt weekly from a golden configuration plus masked production data so testers stop discounting defects as data issues; a production clone for performance, cutover rehearsal, and training. Freeze configuration to a weekly change window with a change board, or the environment drifts faster than it can be tested.
Partner reset. Convert the conversation from effort to outcomes. Ask for a written statement of which of the 17 critical items will be delivered by which date, and for the four-week defect trend - closed versus opened says more than a percentage-complete figure. Put the severity A list and the exit criteria into the contractual governance forum with the CIO present. Where the integrator will not commit, the fallback is a named internal owner per item, not another workshop. If scope must change, negotiate it once, in writing, tied to the go-live decision; sequential renegotiations during a rescue consume the remaining budget faster than the technical work does.