Context
A regional payments fintech with operations across Singapore, Malaysia and Indonesia had grown its multi-cloud footprint (AWS + Azure) to ~SGD 1.7M/year without any of the discipline a CFO eventually demands. Engineering owned spend in name; in practice, nobody owned anything. Untagged resources were 60% of the bill, the Reserved Instance coverage was 6%, and three different teams were spinning up duplicate shared services in different accounts because cross-team visibility was zero.
The board pressed for a number. Engineering pushed back: “we can’t cut without breaking things.” We were brought in for an 8-week independent review and execution.
What we did
1. Tag the world (weeks 1-3)
Every resource got tagged with team, environment, project, cost center. Untagged resources flagged for owner attribution within 14 days or auto-decommissioned. We built a daily report that showed each team lead their own untagged-resource list and the path to fix it. By end of week 3, untagged spend was under 4%.
2. Right-size systematically (weeks 3-5)
Pulled 90 days of utilization data, identified ~140 over-provisioned instances and 30+ idle resources across the two clouds, retired or downsized them. Engineering reviewed every proposed change before execution — no surprises in their on-call.
3. Reserved-capacity strategy (weeks 5-7)
Replaced on-demand spend on the stable production workload with a mix of 1-year and 3-year RIs (AWS) and Reserved Instances + Savings Plans (Azure). Conservative coverage — 65% — with room to expand once we were confident in the new baseline.
4. Tag-driven chargeback (weeks 7-8)
Built a monthly per-team statement so engineering leaders saw their own bill, not just a global number. Three months in, two teams voluntarily decommissioned services they’d forgotten they owned.
Results
- -38% total cloud spend reduction vs. FY24 baseline
- SGD 650K annualised savings, mostly from RI commitments + idle resource reaping
- Zero production incidents attributable to the FinOps work
- First on-time per-team chargeback statement in company history
- 65% → 72% RI coverage after the post-engagement quarterly review
What we learned
Tag hygiene is 70% of FinOps — until every resource has an owner, nothing else moves. The visible scoreboard (per-team chargeback) was the surprise unlock. Engineering teams stopped arguing with finance once they saw their own bill in their own currency.
The work survives because we left behind not just a smaller bill but a quarterly review cadence — the customer’s own FinOps practitioner now runs the review we built, and we’re invited back annually for an independent check.