Cost Optimisation · Featured
A FinOps engagement that cut a Series-B payments platform’s monthly AWS spend by a third without trading away latency or reliability.
Cost Optimisation
32%
Monthly Savings
A Series-B fintech operating across the EU, regulated as an electronic-money institution, processing millions of card-not-present transactions each month. Engineering was 28 people split across payments, ledger, and platform. Cloud sat on AWS in two regions, but cost ownership was diffuse — no FinOps function, no chargeback, no real cost dashboards beyond the default Cost Explorer view. The board wanted a clear answer to "are we burning money on infrastructure that we don’t actually need?" before the next funding round.
The challenge
A fast-growing payments company saw AWS spend spiral to roughly €30k/month, threatening unit economics ahead of a Series-C raise. Compute, NAT, and unmonitored data egress were the three biggest line items, and nobody on the engineering team had a confident view of where the money was actually going.
Approach
Two-week discovery: tagged every active resource, mapped spend to product surface, and produced a single ranked list of the top twenty cost drivers.
Rightsized non-prod fleets first — a low-risk, fast win that recovered roughly 11% of spend before any architectural decisions had to be made.
Negotiated a three-year compute Savings Plan sized to the realistic floor of production workload, plus selective reserved capacity for the OLTP databases.
Reworked S3 lifecycle policies and CloudFront caching for transaction receipts; introduced gateway endpoints to remove avoidable NAT egress.
Stood up a thin FinOps cadence: a weekly anomaly review, a monthly cost-per-transaction metric, and a tagging policy enforced via Service Control Policies.
The solution
A focused FinOps audit followed by surgical rightsizing, a Savings Plan committed at the workload’s honest floor, and lifecycle policies on the long-tail S3 buckets that were quietly accumulating receipts.
The second-order win was governance: a tagging policy enforced at the organisation level, anomaly detection wired into a Slack channel that the platform team actually reads, and a cost-per-transaction metric the CFO can quote in a board meeting without translating from engineering language.
Monthly Savings
Compute Cost
Time to Value
Reflections
The cost story was never really about EC2 — it was about ownership. Once a single engineer owned the weekly cost review and the tagging policy had teeth, the savings stuck. A year on, cost per transaction is still trending down without anyone needing heroics. Production stays boring, the spend stays predictable, and that’s the version of "cost optimisation" that holds up to a board pack.
Continue exploring
Engagements are scoped around outcomes, not hours. The fastest path is a short discovery call—no slides, just questions.
