Right sizing before you commit is the single most leveraged piece of FinOps work you will ever do, because the baseline you carry into a multi year deal becomes the floor you pay against for the life of the term. Sign a commitment sized on a bloated bill and you have not bought a discount, you have funded waste at a contracted rate. This is part of our wider work on FinOps optimization for cloud commitments, and it pairs directly with commitment structuring and sizing advisory. Get the number right here, before signature, and every downstream negotiation gets easier.
Why right sizing before you commit beats right sizing after
Once a commitment is signed, optimization works against you. Cut a workload and you free capacity you have already promised to buy, which can push you toward a shortfall. Right sizing before you commit reverses that pressure. Every instance you shrink, every idle resource you kill, and every overprovisioned database you tune lowers the number you are about to lock in. The savings compound across the full term rather than being trapped behind a commitment you cannot escape.
As of June 2026, the major committed use programs all share the same mechanic. An AWS EDP, an Azure MACC, and a GCP committed use deal each set a dollar floor you must consume or pay for. Right sizing after signature does not lower that floor. It only changes how you fill it. So the cleanup has to happen first.
Find the waste the seller wants in your baseline
Vendors size proposals from your trailing run rate, because a higher run rate justifies a higher commitment and a bigger headline discount. That is not your interest. Before you commit, separate the spend you actually need from the spend that exists only because nobody turned it off. Idle compute, oversized instances, unattached storage, forgotten test environments, and stale snapshots are the usual suspects.
Work from utilisation, not from the invoice. A fleet running at twenty percent average CPU is a fleet half its size waiting to happen. The cleaner your demand picture, the harder it is for a sales team to anchor you on a number built on air.
Turn a clean baseline into negotiating leverage
A right sized baseline does more than lower the commitment. It changes the conversation. When you can show that your committed number reflects real, governed demand, you remove the seller's easiest objection, which is that you are sandbagging. You also protect yourself from punitive ramp assumptions, because your forecast is grounded in evidence rather than optimism.
Pair the cleanup with a coverage view. Decide how much of your steady state you want under commitment and how much you want to leave flexible. We cover that split in detail in our work on coverage and utilisation targets.
A pre commitment right sizing checklist
Run the cleanup in a fixed window so the savings show up in the baseline you take to the table. Terminate idle and zombie resources. Downsize instances that never approach their ceiling. Move cold data to cheaper tiers. Consolidate underused clusters. Remove duplicate tooling. Then rebaseline and forecast from the cleaned figure, not the original one.
Document every change. When the provider argues your number is too low, your evidence is the list of waste you removed and the utilisation that proves the rest is real. That record is worth more at the table than any spreadsheet of list price discounts.
Tooling makes pre commitment right sizing repeatable
Right sizing by hand does not scale across an enterprise estate. Use the native optimization recommendations each provider exposes, then validate them against your own utilisation data before acting. As of June 2026 the major clouds all surface idle and oversized resource flags, but those recommendations are tuned to the provider's view, not yours, so treat them as a starting list rather than a verdict.
Layer a tagging and ownership model underneath the tooling so every resource has an owner who can confirm whether headroom is real or waste. The combination of automated detection and human confirmation is what produces a baseline you can defend at the table rather than a guess you hope holds.
The right sizing mistakes that survive into the commitment
The most expensive mistake is sizing peak as if it were steady state. A workload that spikes once a quarter does not need full capacity reserved year round. Another is treating non production like production, leaving development and test environments running and oversized when they could be scheduled and shrunk.
A third is ignoring data gravity. Storage and transfer costs creep into the baseline quietly and then ride along in the committed number for years. Catch these before signature, because every one of them becomes a contracted cost the moment the deal closes.
How right sizing changes by provider
The principle is constant but the levers differ. On AWS the work centers on instance families, Graviton migration where it fits, and matching savings plans to the cleaned compute profile. On Azure the focus is VM series selection, reservation scope, and trimming the services that quietly accrue under an enterprise agreement. On GCP, sustained use discounts apply automatically, so right sizing interacts with what you then choose to commit.
Whatever the provider, the sequence holds. Clean first, measure utilisation, then decide how much of the cleaned demand belongs under commitment. The commitment should follow the right sized baseline, never lead it.
What good looks like once right sizing is done
A finished right sizing pass produces three things. A demand picture grounded in utilisation rather than the invoice. A documented list of the waste removed, which is your evidence at the table. And a forecast built from the cleaned figure that you can defend in a downside case.
With those in hand, the commitment conversation changes. You are no longer negotiating a discount on a number the seller built. You are presenting a number you built, backed by data, and asking the provider to price it. That is the position right sizing buys you.