An observability platform for 40 fintech microservices
From 60 dashboards nobody opened to 12 SLOs the team acts on — MTTR down 63% in one quarter.
- Client
- Fintech · Bangkok
- Role
- Observability design & SRE advisory
- Timeline
- 10 weeks
- Stack
- Prometheus, Grafana, Loki
-63%
MTTR
12
active SLOs
-78%
night pages
40
microservices
01 — The challenge
Plenty of dashboards, no agreement on what mattered. Incidents took 45 minutes just to localise, and logs were spread across four systems.
02 — Approach
Started from 12 business-critical journeys, defined error budgets, and alerted on burn rate instead of static thresholds. Consolidated logs into Loki and sampled traces at the right boundaries.
03 — Outcome
MTTR fell from 46 to 17 minutes, night pages dropped 78%, and the team gained a shared language for risk.
Stack
Principle
We deleted 60 dashboards down to 12 — not because the graphs were ugly, but because every panel must answer "if this is wrong, how do users suffer?"
Burn-rate alerts instead of thresholds
- alert: CheckoutErrorBudgetBurn
expr: |
(sum(rate(http_requests_total{route="checkout",status=~"5.."}[5m]))
/ sum(rate(http_requests_total{route="checkout"}[5m]))) > (14.4 * 0.001)
for: 2m
labels: { severity: page }Measured results
- MTTR 46 → 17 minutes
- 78% fewer alerts that required no action
- New-engineer onboarding to production insight: 2 days, down from 2 weeks