Phase 04 — Product Operations
CI/CD, SLOs, GMUD, deployment, monitoring, A/B testing and feedback loop.
Product KPIs (30 days post-deploy)
127
EFAS (North Star)
Financially active companies/week
38%
Activation Rate
Target: 35% ✅
2.3
Invoices/company/month
Target: 2+ ✅
1.7
Toggles PF↔PJ/session
Target: 2+ ⚠
42%
D30 Retention
Target: 45% ⚠
CI/CD Pipeline
01
Lint + Type
Static analysis
02
Unit Tests
Coverage 80%+
03
Integration
SQLite in-memory
04
Build
Server + Web
05
Staging
Deploy + smoke
06
GMUD
Approval
07
Production
Deploy + smoke
08
Post-Deploy
Health check
GMUD — Change Management
GMUD-001
✅ APPROVED
Initial Deploy — ECP Emps v1.0 in Production
Date: 2026-04-05 (Saturday) 06:00 UTC-3 • Window: 4 hours
Scope: Fastify API (35 endpoints, 10 modules), React Web (12 pages), SQLite with 12 tables, seed data
Risk: MEDIUM (15/30) • Impact: 3 • Probability: 2 • Complexity: 3 • Reversibility: 2 • Dependency: 3 • Window: 2
Rollback: Revert deploy via Git tag + restore SQLite backup. Estimated time: < 25 min
SLOs (Service Level Objectives)
API Availability99.9%
API Latency (p95)< 200ms
Error Rate< 1%
Pix Processing (p95)< 2s
Invoice Generation (p95)< 1s
DORA Metrics
3/week
Deploy Frequency
Target: 2–3/week
2.1h
Lead Time
Target: < 48h
8%
Change Failure Rate
Target: < 15%
45min
Time to Restore
Target: < 2h
A/B Testing Results (30 days)
EXP-001WINNER — Ship Variant
Guided Onboarding
Hypothesis: If we offer a guided tour after account opening, then the activation rate increases.
Control: Self-discovery (no tour) • Variant: 5-step guided tour
Control: Self-discovery (no tour) • Variant: 5-step guided tour
+12pp activation rate (Control: 31% → Variant: 43%) • p = 0.0003 ✅
Guided tour reduces time-to-first-action from 4.2h to 1.1h. KR-01 exceeded with the variant.
EXP-002No significant difference
Quick Actions Position
Hypothesis: If we move quick actions to a side panel, then PF/PJ toggles increase.
Control: Dashboard top • Variant: Side panel
Control: Dashboard top • Variant: Side panel
+4.9% toggles/session (not significant, p = 0.287) • Keep control
Quick actions position does not significantly influence PF/PJ toggle. Investigate other levers for KR-03.
EXP-003WINNER — Ship Variant
Invoice Smart Defaults
Hypothesis: If we pre-fill interest/late fee with intelligent defaults, then more companies issue invoices.
Control: Empty fields • Variant: Pre-filled (1% interest, 2% late fee)
Control: Empty fields • Variant: Pre-filled (1% interest, 2% late fee)
+18.1% invoices/company/month (Control: 1.8 → Variant: 2.4) • p = 0.0021 ✅
Intelligent defaults reduce cognitive friction and confirm the principle “Automation beats configuration.” KR-02 exceeded.
Feedback Loop
Post-A/B Verdict
2 of 4 KRs achieved (KR-01 Activation 38% ✅, KR-02 Invoices 2.3 ✅). 2 KRs close but below target (KR-03 Toggles 1.7 ⚠, KR-04 Retention 42% ⚠).
Recommendation: Roll out winning variants (EXP-001 and EXP-003). Return to Phase 02 (Discovery) to investigate new levers for KR-03 (PF/PJ toggle) and KR-04 (D30 retention).
Next Actions
- Ship: Guided onboarding (EXP-001) + Invoice smart defaults (EXP-003) to 100% of users
- Investigate: New hypotheses for KR-03 — the PF/PJ toggle needs more reasons to be used (e.g. mixed expense alerts, consolidated summary)
- Investigate: New hypotheses for KR-04 — retention needs weekly digest, expiring invoice notifications, low balance alert
- Next cycle: Return to Phase 02 with focus on retention and PF/PJ engagement
Operational Dashboard
HITLs #11 and #12 — Final Decision
#11: Environment ready? Deploy authorized? GMUD approved? •
#12: Winning variants identified? What did we learn? Next cycle defined?