The Go or No-Go Decision Matrix
Estimated read: 3 min
The Migration Playbook introduced the go/no-go decision matrix as the artifact that turns a cutover call from a gut decision made under pressure into something defensible and auditable. Here’s what it actually looks like filled in, walking into this capstone’s war room:
| Criterion | Source | Status |
|---|---|---|
| Reconciliation status | Nightly hash-diff, all migrated table pairs | Green – zero open mismatches as of last night’s run |
| Rollback readiness | Rehearsed backfill drill against a deep clone | Green – last drill reconciled clean in 40 minutes |
| Business sign-off | Trust-gate sign-off from finance and ops report owners | Yellow – finance signed off; ops still validating two reports |
| Compute headroom | Cluster sizing vs. peak parallel-run load | Green – provisioned for 2x observed peak |
| Cutover gap tolerance | Estimated write volume during the decision window | Green – under 10 minutes of transactions at current load |
Four columns, four sources, four statuses – and the row that matters here is the one that isn’t green. This is what a gate, not a score, means in practice: this matrix is never averaged. A 4-of-5 green matrix is not “80% ready to go.” One unresolved criterion, however small it looks next to four green rows, is enough to hold the cutover – because the criterion that’s yellow is a business sign-off, and going live over an open sign-off gap converts a process risk into a trust problem the moment something – anything – goes wrong on a report ops hasn’t finished validating.
Treat every non-green row as a hold, never as a number to net against the green rows. A matrix that lets four strong criteria outvote one weak one isn’t a decision tool – it’s a way to talk yourself into a cutover you’d have blocked if the weak criterion had been the only one in front of you.
Notice, too, what each row’s source column is doing: every status traces back to a rerunnable artifact from earlier in this part, not a verbal assurance from someone in the room. Reconciliation status comes from the audit table behind the dashboards you built. Rollback readiness comes from a timestamped drill log, not a claim that “we tested that once.” Compute headroom comes from actual parallel-run telemetry, not a sizing guess from the original project plan. A go/no-go matrix with a source column that says “team confidence” for half its rows isn’t a matrix – it’s a vote, and votes are exactly what this artifact exists to replace.
For this capstone, the ops sign-off gap closes overnight: the two outstanding reports get validated, and every row shows green by the time the next lecture’s 48-hour window opens. That’s the version of this exercise where the matrix does its job quietly – the harder version, where the matrix catches something real mid-window, is exactly what the next lecture walks through live.
| ← Previous: The Scenario: 50TB Oracle EDW, 500 Procs, 200 Reports, $50M Line | Next: The 48-Hour War Room: Live, No-Go, or Rollback → | |:—|—:|