A good Azure cost optimization assessment answers three questions in order: where is the money going, how much of it is waste, and what is the fastest path to remove the waste without hurting the workload. Run in that sequence, a single assessment routinely surfaces double-digit percentage savings before you negotiate a single rate. The discipline is doing the steps in the right order so you never commit budget to resources that should have been shut off.
This article is part of our Azure cluster. For the full landscape, start with the complete guide to Azure cost optimization, the pillar this piece links up to. An assessment is the See step of our See, Cut, Lock, Run method made concrete: you cannot cut what you have not first measured and attributed.
Export your Azure cost data at the resource level, not the subscription summary. Use Cost Management exports to a storage account, or the FOCUS-formatted export if available, so every line carries a resource ID, resource group, tags, meter, and the amortized cost. Amortized cost spreads reservation purchases across the term so you see true unit economics, not a lumpy upfront charge.
Step 1: Scope and baseline the spend
Pull at least three months of amortized cost broken down by subscription, resource group, service, and tag. Three months smooths out one-off spikes and reveals the steady run rate that commitments will eventually cover. Build a simple Pareto view: the top services and the top resource groups almost always account for the large majority of the bill. Compute, SQL, storage, networking, and increasingly Log Analytics tend to dominate. That ranking tells you where the assessment effort pays off and where it would be busywork.
Tag coverage is the gate on everything that follows. If a large share of spend is untagged, you cannot allocate it to a team or product, and you cannot tell waste from legitimate cost. Measure tag coverage as a percentage of spend, not a percentage of resources, because a handful of expensive untagged resources matters far more than many cheap ones.
Step 2: Find the obvious waste first
Idle and orphaned resources are the cheapest savings because removing them carries almost no risk. In an Azure assessment we look for unattached managed disks, orphaned public IP addresses, idle load balancers and gateways, empty App Service plans, stopped-but-not-deallocated VMs that still bill for compute, and old snapshots and backups that no one owns. Azure Advisor surfaces many of these, and we cover its strengths and blind spots in how to use Azure Advisor for cost recommendations. The point of doing this step before rightsizing is simple: there is no reason to carefully rightsize a VM you are about to delete.
Want the assessment run for you, end to end?
Our Azure cost audit profiles every subscription, quantifies the waste, and models the rightsizing and commitment moves with a ranked savings plan. On the performance model, you pay only from realized savings. No savings, no fee.
Book an Azure cost audit →Step 3: Rightsize the survivors
With the dead weight gone, size the resources that remain to their real workload. Read at least two weeks of CPU, memory, and IO, then match the VM size to the high-percentile utilization with headroom rather than the peak-of-peaks. The full procedure lives in how to rightsize Azure virtual machines, and the same logic extends to databases, disks, and App Service plans. Rightsizing is where the durable savings come from, because a smaller resource is cheaper every hour forever, and it shrinks the baseline you will later commit to.
Step 4: Model commitments on the clean baseline
Only now, on a rightsized and de-cluttered estate, do you model rate discounts. The two instruments are Azure Reservations and the Azure savings plan for compute, and choosing between them is its own decision covered in Azure Reservations vs Azure savings plan for compute. Layer Azure Hybrid Benefit on top where you have eligible Windows Server or SQL Server licenses. Buying commitments last is the single most important sequencing rule in the assessment: commit first and you lock in the oversized estate for one or three years.
| Assessment step | What you look at | Typical outcome |
|---|---|---|
| 1. Baseline | 3 months amortized cost by service, RG, tag | Pareto view + tag coverage gap |
| 2. Waste | Idle, orphaned, stopped-not-deallocated | Low-risk immediate cuts |
| 3. Rightsize | 2+ weeks utilization on survivors | Durable per-hour savings |
| 4. Commit | Reservations, savings plan, Hybrid Benefit | Rate cut on clean baseline |
| 5. Rank | Savings vs effort vs risk | Prioritized action plan |
Step 5: Rank by savings, effort, and risk
An assessment that ends in a list of two hundred recommendations gets ignored. Score each move on annual savings, implementation effort, and risk to the workload, then sequence the quick, safe, high-value items first to build momentum and fund the harder work. Deleting orphaned disks is high value and near-zero risk; consolidating a database tier is higher value but needs a change window. The deliverable is a plan a team can execute, not a data dump.
The tooling and meter names referenced here reflect Azure Cost Management as of May 2026. Verify current export formats, Advisor categories, and commitment terms in Azure documentation before acting, as the platform evolves. For the wider context on reading the bill, see understanding Azure Cost Management and billing.
The Azure Cost Optimization Field Guide includes the assessment worksheet, the waste checklist, and the savings-ranking template we use on engagements. It is the downloadable companion to this article.
The short version
Run the assessment in order: baseline the spend with tagged, amortized data; clear the obvious waste; rightsize what survives; model commitments on the clean baseline; then rank every move by savings, effort, and risk. For the fast wins to act on first, see the Azure cost optimization checklist. When you want the whole assessment run and the plan executed for you, that is exactly what our Azure cost optimization service delivers.