Lesson 3 of 5
Quality gates
11 min
Listen to this lesson
SAP Activate uses quality gates at the ends of its phases: checkpoints where the project confirms that deliverables are complete before moving on. In a brownfield program, the most valuable gates sit at the conversion cycles, because each conversion commits time, people and systems. A good gate has criteria anyone can check, evidence that is stored, a named decision maker, and only three outcomes: pass, conditional pass, or fail.
Contoso's gates
- G1: Ready to convert CMDSandbox stable; Maintenance Planner plan approved; blocker list closed or owned
- G2: Ready to convert CMQTransports complete; readiness ATC clean in scope; finance Customizing ready
- G3: Ready for Mock 1CMQ test cycles complete within thresholds; close passed; runbook draft
- G4: Ready for dress rehearsalMock 2 fits window; fallback measured; reconciliations at zero
- G5: Ready for productionDress rehearsal passed; final readiness review complete
Writing measurable criteria
| Weak criterion | Measurable criterion | Evidence |
|---|---|---|
| Custom code is mostly done | Readiness ATC run on CMD shows 0 open errors for in-scope objects | ATC result, dated |
| Finance looks fine | All reconciliation rows for 1710 and 1810, ledgers 0L and 2L, at zero difference | Reconciliation file signed by controllers and Dan |
| Testing is going well | Regression pass rate at least 95%; 0 open critical and at most 5 open high defects, each with a workaround | Test and defect reports from SAP Cloud ALM |
| Data is clean | SI-Check shows no return code 8 or 12 on a production copy taken this month | SI-Check result list |
| Cutover is ready | Dress rehearsal fitted 60 hours with at least four hours of contingency | Rehearsal timeline from the runbook |
From opinion to evidence. Thresholds are Contoso's; set your own with the steering group.
Thresholds are a program decision, made before the gate and not at it. Priya proposes the testing thresholds, Dan the finance ones, Luis the cutover ones; the steering group approves them. Changing a threshold at the gate to let a cycle pass is the most common way gates lose their meaning. If a threshold turns out to be wrong, change it openly, record why, and apply it to the next gate.
Pass, conditional pass, fail
- Are all criteria met with evidence?Yes → pass · No → next question
- Is each unmet criterion contained, with a workaround or a plan that does not threaten the next cycle?Yes → conditional pass: each condition gets an owner, a date and a check at the next gate · No → next question
- Can the gap close before the next cycle must start?Yes → fail now, re-gate on a set date · No → fail and re-plan the cycle dates
A conditional pass is useful and honest when the conditions are specific and tracked. It becomes dangerous when conditions pile up unreviewed. Contoso's rule: every condition has an owner and date, appears on the next gate's checklist as a criterion, and no gate may pass with more than three open conditions carried from earlier gates.
Worked example: G4 after Mock 2
| G4 criterion | Mock 2 result | Status |
|---|---|---|
| Conversion fits the 60-hour window | 58.5 hours | Met |
| Fallback duration measured | 9 hours | Met |
| Finance reconciliations at zero for 1710 and 1810, ledgers 0L and 2L | Zero, signed by Dan | Met |
| Contingency at least 4 hours | 1.5 hours of float | Not met |
Contoso’s G4 scorecard. One criterion is unmet, but it is contained and has a plan.
The gate chose a conditional pass. The unmet criterion did not threaten the dress rehearsal itself; it threatened go-live if left alone. So the condition was specific: Sam owns the SUM tuning, Dan owns parallel reconciliation, and “dress rehearsal finishes with at least 4 hours of contingency” became a G5 criterion. The dress rehearsal met it, and the condition closed with evidence. In SAP Activate terms, Contoso’s G1 to G5 sit inside the Explore, Realize and Deploy phases and feed the phase-end quality gates rather than replacing them.
Evidence in one place
Evidence lives where it is produced, but the gate needs links to all of it: test runs and defects in SAP Cloud ALM, ATC results, SI-Check results, reconciliation files, SUM logs, rehearsal timelines and signed approvals. Grace's PMO keeps a gate pack per gate with the criteria, the status, a link to each piece of evidence and the decision record. Auditors and the steering committee read the same pack.
What would you do?
At G3, the regression pass rate is 93% against a 95% threshold. The failing tests are in one area: custom output forms for delivery notes, with a workaround (standard forms) agreed by Maria. Everything else meets its criteria. Mock 1 is in ten days. What should the gate decide?
- Tie gates to conversion cycles; each gate commits time, people and systems.
- Write criteria as checkable facts with thresholds approved before the gate.
- Outcomes: pass, conditional pass with owner and date, or fail; limit carried conditions.
- Keep a gate pack with criteria, linked evidence and the decision record.
Check your understanding · not graded
Lesson check
1. When should gate thresholds be set?
2. Which is a measurable gate criterion?
3. What must every condition in a conditional pass have?
4. Why limit the number of conditions carried between gates?
5. At G4, Mock 2 fit the window and reconciled, but contingency was 1.5 hours against a 4-hour minimum. Which outcome fits?