Eight years of FinOps for a European payments platform

The client runs a point-of-sale product on AWS. Payment platforms carry an unusual constraint: cost optimisation cannot come at the expense of reliability or latency, because both are the product. When a merchant’s till depends on your platform, an outage isn’t degraded service. It’s a queue of customers who can’t pay, in every shop and café running the product, all at once.

The problem

That constraint rules out most of the standard cost-cutting playbook. You cannot switch things off and see who complains. You cannot resize aggressively and let performance tell you when you’ve gone too far. You cannot treat spare capacity as pure waste, because some of it is headroom the platform genuinely needs for the moments that matter. Every optimisation on a payments platform has to be argued for twice, once on cost and once on risk, and the second argument is the one that counts.

There’s a second, subtler problem: drift. Cloud estates decay towards inefficiency by default. Reservations expire and lapse back to on-demand rates. Workloads change shape as the product evolves, leaving yesterday’s sizing behind. New services accumulate around the edges. An estate optimised once and left alone typically gives back much of the gain within a year, which is why one-off cost reviews photograph the problem rather than solve it.

What we did

Cloudreason has managed reservation strategy and spend monitoring for the platform for eight years, and works directly inside the product’s infrastructure, making the technical changes required to keep cost efficient rather than only recommending them. Engineering capability, not just analysis. On a platform this sensitive, that distinction matters: a recommendation handed over a fence transfers the risk to whoever implements it, while an engineer who owns the change also owns its safety.

In practice, the engagement runs as a continuous discipline rather than a series of exercises. Every optimisation is weighed against its performance risk before it’s made. Changes land deliberately, in step with the product team, and are watched after they land. Commitment coverage is kept continuously current, planned around the platform’s real usage rather than revisited in occasional bursts, so the estate never drifts far from its efficient shape, and reliability is never traded for a discount.

The results

The engagement is now in its eighth year, making it Cloudreason’s longest continuous client relationship.

The longevity is the point, not a footnote. Eight years of continuous attention is what keeps a platform efficient for its whole life rather than for the quarter after a cost review: through product changes, traffic growth and several generations of AWS pricing instruments. The platform has stayed fast, stayed reliable and stayed efficient, simultaneously, for eight years. That is what FinOps looks like when it’s treated as an operating discipline instead of a project.