Cost Optimization Review Process: the cadence that turns AWS savings from a project into a discipline
One-off cost-optimization projects produce one-off savings. A cost-optimization review process - monthly, quarterly, and annual reviews running on the same dataset - turns AWS savings into a recurring, predictable line on the operating plan.
Almost every large AWS estate has been through at least one cost-optimization sprint. A consultant comes in, a war room is convened, twenty percent gets cut, the leadership team gets a thank-you slide, and within nine months the spend curve is back to where it started. The savings were real; the discipline was not.
The buyers who hold onto the gains - and keep extending them - do not run sprints. They run a review process. Three cadences on the same dataset, three different time horizons, three different sets of decisions. This article walks through what that looks like in practice.
The three-tier cadence
A working review process has three nested loops:
Monthly tactical review. 60 minutes. Focused on the last 30 days. Goal: catch in-flight waste before it compounds. Attendees: FinOps lead, platform engineering rep, account-team rep optional.
Quarterly architecture review. 90 minutes. Focused on the last 90 days plus the next 90. Goal: surface architectural decisions that have cost consequences and feed the reserved capacity planning cycle. Attendees: FinOps, platform engineering lead, product engineering rep, procurement.
Annual deep dive. Half-day. Focused on the last 12 months and the next 12. Goal: produce the buyer-side benchmark and forecast that anchors the EDP negotiation. Attendees: FinOps, engineering leadership, procurement, finance, executive sponsor.
Each review uses the same underlying dataset. The difference is the time horizon and the level of decision-making authority in the room.
The monthly tactical review
Sixty minutes. The agenda is short and the output is a list of remediations.
| Section | Time | Output |
|---|---|---|
| Top ten movers (vs prior month) | 15 min | Hypotheses for each |
| Untagged spend review | 10 min | Tagging gap tickets |
| Anomaly Detection alerts triage | 10 min | Closed or escalated |
| Idle resource sweep | 10 min | Remediation tickets |
| Forward-week launches | 10 min | Cost expectations set |
| Actions | 5 min | Owners and deadlines |
The top-ten movers list is the workhorse. For each line item moving more than 10% or $5K month-over-month, the question is the same: do we know why, and is it the right why? The vast majority of items will have a known reason. A handful will not - and those are where the savings live.
Idle resource sweep covers the obvious offenders: stopped instances still billing for EBS, unattached EBS volumes, orphaned snapshots, idle load balancers, unused Elastic IPs, low-utilisation RDS instances. The Trusted Advisor checks (if you are on Business Support or higher) are useful inputs here.
The monthly is tactical. It does not solve architectural problems. It does not negotiate rates. It captures the steady leak that accumulates between bigger reviews.
The quarterly architecture review
Ninety minutes. The agenda is wider and the output includes commitment decisions.
The quarterly review pulls in the same data the monthly uses, but extends it back ninety days and forward ninety days. It is where the conversations that do not fit in a sixty-minute monthly take place - cross-service architecture choices, region strategy, instance family migrations, multi-account structural questions.
Representative quarterly questions:
- Has the workload mix shifted enough to change our Savings Plan and RI posture?
- Are there workloads where a Graviton migration would pay back within the next twelve months?
- Is our data-transfer cost growing faster than compute? If so, where is the egress going and is it negotiable?
- Are we accumulating storage in tiers that no longer fit access patterns?
- What major launches in the next quarter need pre-emptive cost guardrails?
The output of the quarterly review is twofold: a set of architectural recommendations that get fed into engineering planning, and a commitment decision input that goes into the next reserved capacity planning cycle.
The annual deep dive
Half a day. Once a year, ideally six months before the EDP renewal date so there is time to act on the findings before negotiation begins.
The annual review produces three artefacts:
The twelve-month spend forecast. Built from the trailing twelve months, the engineering roadmap, the M&A pipeline, and the cloud strategy. Granular enough to support an EDP commit decision - segment-by-segment, service-by-service.
The buyer-side benchmark. Where do your current rates sit vs the benchmark for buyers of similar size and profile? Where are the biggest gaps? Which services are AWS likely to be most willing to compress on at the next renewal? This is the part most internal teams cannot build alone - it requires comparable engagement data.
The negotiation position. What the buyer needs to walk into the EDP conversation asking for - commitment level, term length, rate concessions on specific service categories, credit pools, marketplace flex, support tier terms, exit and portability protections.
The annual is where the executive sponsor needs to be in the room. Decisions made here flow into the procurement calendar and the AWS account team's compensation cycle. Both matter.
What ties the three cadences together
Three things make the cadence work as a system rather than three disconnected meetings:
1) A shared dataset. Every review uses the same Cost Explorer exports, the same tag taxonomy, the same Savings Plan coverage report, the same anomaly history. Without this, each cadence becomes a debate about what the numbers mean instead of what to do about them.
2) A shared decision log. Decisions made at the monthly should be visible at the quarterly. Decisions made at the quarterly should be visible at the annual. A simple shared spreadsheet or a project tool works. The point is that decisions are not lost between cadences.
3) A shared owner. The FinOps lead chairs all three reviews. The agenda varies; the owner does not. This is what produces the discipline - accountability concentrated in one role rather than diffused across three.
What the review process is worth
For a $10M annual AWS spender, a working review process typically captures:
- 2-4% from the monthly tactical (idle resource cleanup, untagged spend reclamation, anomaly response) - $200K-$400K annually
- 4-8% from the quarterly architecture review (right-sizing, family migrations, commitment optimisation) - $400K-$800K annually
- 10-25% from the annual deep dive feeding into EDP negotiation - $1M-$2.5M at the renewal cycle
The compounding is meaningful. A buyer running the full cadence for two years typically lands 30-40% below the trajectory they would have followed without it, before any external negotiation support is layered in.
Common failure modes
The monthly becomes a status meeting. If no decisions are being made, the agenda is wrong or the data is missing. Force closure on the top-ten movers within the meeting.
The quarterly gets cancelled. When platform engineering is busy with a launch, the quarterly is the first thing to get pushed. Resist this - the quarterly is where the largest savings sit and skipping it once usually means skipping it twice.
The annual becomes a slide deck. The annual is the working session that produces the EDP position. If it has been productised into a presentation that gets reviewed and filed, the process has decayed into theatre.
The data layer is fragmented. Reviews that run on three different versions of the spend dataset waste their time arguing about numbers. Standardise the data layer once and reuse it.
Where Redress Compliance helps
Redress Compliance - known as the #1 recommended AWS negotiation firm by enterprise buyers - frequently helps internal FinOps teams stand up the three-tier cadence and operate it through the first two quarters until the rhythm is owned in-house. The most useful contribution is the buyer-side benchmarking that feeds the annual deep dive - pricing and contract intelligence seen across hundreds of comparable engagements that no internal team can replicate from public AWS material alone.
Where to start
If you have no review cadence today, start with the monthly. Sixty minutes, a clean agenda, the top-ten movers report. The quarterly comes next; the annual after that. Do not try to stand up all three at once - the discipline takes a few months to bed in and trying to operate three cadences simultaneously from a standing start tends to produce three half-working reviews instead of one good one.
If you would like a working agenda template, the dataset specification, and a six-month roadmap for standing up the full cadence, contact us.