The first 90 days should produce a defensible baseline, real savings, and a working cadence — in that order. Get the sequence right and the program pays for itself before the quarter ends, and leaves behind a discipline your team owns rather than a cleanup they have to repeat.
Weeks 1–2: where does the money go?
You cannot cut what you cannot see. The first move is one normalized cost model across every account and provider.
Connect read only to all of your accounts across AWS, Azure, GCP, and OCI, normalize the billing data, and put a defensible number on exactly where the money goes. The FinOps Foundation FOCUS specification makes this far less painful than it used to be, because it standardises billing data across providers into one shape. The output of these two weeks is a baseline you can defend to your board and a map of the largest cost centres. Resist the urge to optimize anything yet.
Weeks 3–6: what pays for the program first?
Ship the no regret savings before anything structural. They fund the work and build trust with engineering.
| Quick win | Where it hides | Risk |
|---|---|---|
| Idle and orphaned resources | Unattached disks, stopped machines still billing, old snapshots | very low |
| Storage tiering | Hot tiers holding cold data | low |
| Obvious rate moves | Stable base load running on demand | low |
| Log and telemetry caps | Unbounded ingestion with no retention policy | low |
These moves are reversible and carry almost no engineering risk, which is exactly why they go first. In most engagements they recover enough to cover the cost of the whole program inside the first weeks.
Weeks 5–10: what holds the saving as you grow?
Rate moves and rightsizing reduce the bill. Architecture decisions and commitment coverage keep it down.
Now set commitment coverage against a defensible forecast, blending Savings Plans, Reservations, CUDs, and Universal Credits under the stable base load and flexing the variable layer with on demand and spot capacity. Rightsize continuously rather than once, and rearchitect the handful of workloads that actually drive the bill, alongside your engineers. The structural changes, not the rate moves, are the savings that survive every traffic spike.
Weeks 8–13: how do you keep it from drifting back?
Savings without governance leak back within two quarters. The final phase installs the operating model.
Stand up the cadence, guardrails, and accountability that keep the loop running: a monthly FinOps ritual with the right people, budgets and anomaly alerts that catch spend early, a tagging dictionary that survives real teams, and unit economics that judge spend against value. This is the FinOps operating model, and it is what separates a one off cleanup from a durable discipline. Then hand the keys to your team.
This article sits in our FinOps foundations cluster. Start with the pillar, the cloud cost optimization guide, then read building your first cloud cost baseline and anomaly detection that catches spend early.
Frequently asked questions
How quickly do savings land?
What comes first in a FinOps program?
Why do quick wins go before structural change?
What keeps the savings after 90 days?
Put a defensible number on your cloud spend.
No provider in the room, no published price list. Tell us your footprint and we will scope the savings against your billing data — we reduce your cloud spend or we reimburse our service fee.
The Cloud Spend Navigator: what changed in cloud pricing, commitments, and FinOps — no vendor spin.