Azure Advisor scans your resources and produces cost recommendations: shut down or resize underused virtual machines, buy reservations or an Azure Savings Plan for steady workloads, and remove idle resources such as unattached disks and unused public IP addresses. The recommendations are genuinely useful and a strong starting point, but Advisor is an advisor, not a decision. Its rightsizing logic reads a default lookback window of recent CPU and memory metrics, so it can miss a monthly or quarterly batch peak and flag a correctly sized VM as underused. Its reservation recommendations assume your recent run rate continues and do not know your forecast, your planned migrations, or the reservations you already hold, so acting on them blind can overcommit. The discipline is to trust the direction and verify the specifics: extend the lookback or check the workload's real cycle, confirm a resize keeps you inside existing reservation coverage rather than stranding it, and size new commitments to a defensible forecast rather than to Advisor's extrapolation. Verified picks bank reliably; raw picks occasionally page you at 2am or lock in a commitment you will not use.
Here is what Advisor measures, where its defaults mislead, and the checklist that makes its cost picks safe to act on.
What does Azure Advisor actually recommend?
Advisor produces cost recommendations in a few families. Rightsizing or shutdown for virtual machines it judges underused on recent CPU and network metrics. Reservation and Azure Savings Plan purchases for compute it sees running steadily, with an estimated saving against pay as you go. Idle resource cleanup for unattached managed disks, unused public IP addresses, and idle gateways. And some database and storage specific suggestions. Each comes with an estimated saving, which is a projection based on recent usage continuing, not a promise. The estimates are a useful way to rank where to look, but the saving is only real once you have verified the recommendation against the workload's true behaviour.
Where do Advisor's defaults mislead?
Three gaps cause most bad picks. The lookback window is short by default, so a VM that runs a heavy month end or quarter end job can look underused on the other days and be flagged for downsizing. Memory is not always measured without the right diagnostics, so a memory bound workload can be judged underused on CPU alone. And Advisor does not know your commitment position: it can recommend resizing a VM that is currently covered by a reservation, which strands that reservation and raises net cost even as the on demand rate falls. None of these are Advisor being wrong about what it measures; they are Advisor not seeing the context only you have.
How do you verify a rightsizing pick before acting?
Run each pick through four checks. Does the lookback cover the workload's real cycle, including any monthly or quarterly peak, or do you need to extend the window. Is memory actually measured rather than assumed. Does the target size keep the VM inside existing reservation or Azure Savings Plan coverage, or does it strand a commitment. And does the application have a latency or headroom requirement the metrics do not capture. A pick that passes all four is safe to bank. One that fails any is a hypothesis that needs a closer look before you touch production. Roll accepted changes out behind monitoring with a fast rollback.
How do you verify a reservation recommendation?
Advisor sizes a reservation to your recent run rate, which is the wrong anchor if your estate is changing. Before you buy, build a defensible forecast: account for planned migrations on or off Azure, seasonality, and any decommissioning already scheduled. Check what reservations and savings plans you already hold so you do not double cover the same capacity. Choose the commitment instrument deliberately, since the Azure Savings Plan trades a lower discount for flexibility across VM families and regions while a reservation gives a deeper discount for a specific size, a trade worked through in the sibling comparison linked below. Size coverage to the stable base of your forecast, not to the peak, so you carry utilisation risk you can defend.
How does this fit a broader Azure program?
Advisor points at opportunities; the program turns them into a durable reduction. Verify rightsizing against commitments first so you never strand a reservation. Pair downsizing with Azure Hybrid Benefit and Dev Test pricing where eligible, since those change the math independently. Read the amortised cost export as the source of truth rather than the Advisor summary. And review on a cadence so picks are revisited as workloads change. Advisor is one input to that loop, not a substitute for it.
A worked example
A Fortune 500 manufacturer accepted a batch of Azure Advisor rightsizing recommendations across a fleet and saw cost fall, then hit intermittent slowness on a finance service whose monthly close drove a peak the default lookback never saw. Extending the lookback and checking memory corrected several picks, and cross referencing against existing reservation coverage stopped two resizes that would have stranded a commitment. The verified set still delivered most of the projected saving, and it fed the wider program that left the company materially lighter on Azure spend with no repeat incident. Figures are verified against billing data and anonymised.
Frequently asked questions
Is Azure Advisor accurate for cost savings?
Should I act on Azure Advisor reservation recommendations directly?
Can Azure rightsizing strand a reservation?
Turn Advisor output into banked Azure savings
We help Azure teams read Advisor critically, verify each pick against real workload context and existing commitment coverage, and roll changes out without regressions, across AWS, Azure, GCP, and OCI. We take zero provider commissions and answer only to you, on a Fixed Fee or a no risk Gainshare basis, with a simple guarantee: we reduce your cloud spend or we reimburse our service fee. Download the Azure cost playbook, and read the full method below.
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.