By Yair Knijn · February 4, 2025

The FinOps hire every VP Engineering makes 18 months too late

Most VP Engineering tell themselves the same thing about the cloud bill: it is growing, but it is not yet a problem. The flaw is the assumption that FinOps ownership is something you buy when the spend starts to hurt. By the time it hurts, the expensive decisions are already a year and a half behind you, made by nobody, owned by nobody, and baked into the architecture.

FinOps is not a fire brigade you call after the building is alight. It is a control you install while the building is still cheap.

The signal you mistook for "not yet a problem"

The bill is a lagging indicator. The leading indicators showed up much earlier and you have seen all of them: a region spun up for one customer and never decommissioned, a Savings Plan that lapsed because the buyer changed teams, three accounts where untagged is the largest cost bucket, a forecast finance builds by padding last quarter because nobody owns a real one. None of those is a board topic on its own. Each is one accruing interest. You read "spend is up but margins are fine" as permission to wait. It was the last cheap moment to begin.

What un-owned commitments actually cost

Commitment discounts are where the waste calcifies hardest, because the mistakes are time-locked. Buy a one-year Savings Plan against the wrong workload and you are wrong for a year, paying on-demand for what you should have covered. Commit three years to an instance family you outgrow in month four and you carry that liability for the remaining thirty-two. Without an owner watching utilization weekly, nobody catches the gap or the over-commitment until renewal forces it.

Tagging is worse, because you cannot retroactively allocate a cost you never tagged. Eighteen months of untagged spend is eighteen months you can never show a unit economic for. That history does not come back when you finally hire someone. It is gone.

Why a tool without an owner is just a dashboard nobody reads

The reflex move is to buy a cost platform and call it FinOps. It surfaces the anomalies well enough. Then the alert lands in a shared channel, three engineers each assume one of the other two will look, and the idle NAT gateway keeps billing for another month. A tool tells you what is wrong. It cannot decide that fixing it is worth an engineer's afternoon, negotiate the tradeoff with the team that owns the workload, or hold a commitment strategy steady across four quarters. The cheapest savings are almost always the line items nobody was assigned to. A dashboard does not assign anyone; a person does.

The maturity curve, and where late hires actually start

The FinOps Foundation frames maturity as Crawl, Walk, Run, and the order is not negotiable: visibility first, then planning discipline, then optimization. You cannot optimize what you cannot yet see. Industry surveys put only a small minority of organizations at the "Run" stage; most are stuck earlier, still fighting to understand their spend rather than reduce it.

Here is the trap buried in that order. The owner you hire late does not start at Walk because you delayed. They start at Crawl, building the visibility foundation that should have existed six quarters ago, while the bill compounds around them. Waiting does not move you up the curve; it just makes you start climbing later, with a heavier backlog.

Funding the role before the board asks the question

The honest framing for the budget conversation is not "we need someone to cut the bill." It is that an owner is the cheapest insurance you can buy against decisions that turn permanent the moment they are made. The return opens long before the spend is painful, because the savings that matter most are the ones you prevent. You rarely claw the rest back in full.

To surface the un-owned decisions before they calcify, a Cloud Horizons workspace maps commitment coverage, utilization drift, and untagged spend across every account in one view, so the owner you hire starts at Walk instead of Crawl. See how the FinOps workspace works.