The engineer says: everything’s optimized.
The bill keeps growing.
Nobody’s lying. Nobody was looking.
That’s how FinOps engineers happen. Sometimes that’s the right call. Sometimes it isn’t.
A colleague went through a FinOps interview recently.
First hour — tool enumeration. Like a certification exam, not a conversation about money.
Then the second half: cost-per-request thinking, forecasting, shared cost allocation across teams. That part was interesting.
But the real signal was this: the company is large, spends a lot, and needs a dedicated person to watch where the money goes.
Nothing wrong with that. It’s just a diagnosis.
In 16 years of DevOps I’ve never had a “FinOps engineer” on the team. But someone had to think about the money.
I remember a call where we were planning a PoC. The plan: spin up an RDS instance at $600/month. For a proof of concept. With minimal load. For months.
I asked: “Do we actually need that size? Low traffic, but it’ll run for a while — can we go smaller?”
The team laughed. I laughed too, to fit in. Nobody answered. Not even the tech lead.
We launched it. Ran for a few months until the PoC wrapped. Money gone. Question never answered.
Another one: I calculated the cost of identical API workloads on AWS and GCP. Same traffic, same code. The difference was 2x. Nobody on the team knew — because nobody had checked.
A colleague — senior DevOps, FinOps not his specialty — opened Cost Explorer and looked at the top spend. Biggest line item: CloudTrail. Which costs almost nothing. Turns out, in some accounts it had been set up “for reliability” — logs written to two regions. The cross-region traffic was generating the bill. Fixed it the AWS-recommended way: one region, S3 sync. Saved $720K/year.
AWS literally teaches this on their exams: “you can do it either way, but this way is cheaper.” Nobody remembers advice that comes for free.
Back to FinOps.
Another pattern I’ve seen more than once: a company asks one of their engineers to “keep an eye on costs.” He finds issues, adds them to the backlog. A few months later, the manager asks: “What did we save?” Answer: “There are tasks in the backlog.”
Not the engineer’s fault. Structural problem:
FinOps without authority = a list of recommendations nobody prioritizes.
When does a dedicated FinOps hire actually make sense?
The math is simple. A FinOps engineer costs minimum ~$100K/year. To justify that, they need to find and save a comparable amount.
If your infrastructure spend is under $300–500K/year, a dedicated hire probably won’t pay off. Better approach: engineers with cost mindset, plus periodic audits.
But money isn’t the only factor. The more important question: does this person have real authority to make changes?
FinOps that “recommends” is an expensive backlog manager.
FinOps that can say “we’re turning this off” — that’s a different story.
What if you’re not at that spending level yet?
Ask your own engineers to look — they know the infrastructure better than any outside specialist. But they need time and authority, or you’ll just get another backlog.
Or bring in someone for a one-time audit — without pulling your team off their work or creating a permanent role for a periodic task. On average, a first audit finds around 30% savings — one recent engagement came in at 50%. No provider switch, no downtime. For context: Flexera puts the industry average at 42% of cloud spend wasted.
“It works” ≠ “it works at a reasonable cost” ≠ “it’ll be cheap tomorrow.”
Honest question: do you know what each service costs you right now?
I audit IT infrastructure — find not just what’s broken, but what’s quietly draining your budget.
If your AWS/GCP bill has line items nobody can explain, let’s talk.
https://itaudit.yushkov.org/