Back to the blog

How do you prepare for an Azure Well-Architected Review?

Published Oct 06, 2026

The short answer

Preparation comes down to three things: read-level access to the subscriptions in scope, an honest inventory of what is actually running, and the right three people booked for the readout. Teams that do those before kickoff get a week of assessment back. Teams that do not spend week one on logistics.

Sort access, inventory, and people, in that order, before week one starts.

Everything else is optional. Those three decide whether the discovery week produces findings or produces a scheduling thread.

Step 1: Sort the access

Read-level access to every subscription in scope, provisioned before kickoff.

This is the single most common delay. Access requests route through a security team that has never heard of the engagement, sit for four days, and eat most of discovery week.

What to line up:

  • Reader role on the subscriptions in scope, for the assessing team
  • Azure Advisor visibility, which comes with Reader
  • Cost Management read access, which does not always come with Reader and is frequently the one that gets missed
  • Microsoft Entra ID directory read, for the identity portion of the security pillar
  • A named internal contact who can approve an access exception the same day

If your security posture does not allow external Reader access, say so at scoping rather than at kickoff. The assessment can be run in a screen-share model instead. It costs more and the findings are weaker, but it is a real option and it is far better than discovering the constraint in week one.

Step 2: Inventory what is actually running

The architecture diagram and the live inventory are different documents in every estate, and the gap between them is often itself a finding. What you want here is the live one.

Useful to have ready:

  • A current subscription and resource group list, with an owner name against each
  • Which workloads are production, which are not, and which ones people are unsure about
  • Any workload nobody wants to touch, and why
  • Recent incidents, with rough dates
  • What changed in the last 12 months: migrations, acquisitions, new AI or data workloads

The "unsure about" list matters more than it looks. Orphaned environments that nobody will claim are consistently among the fastest cost wins, and they only surface when somebody is willing to say out loud that they do not know what a resource group is for.

Step 3: Book the right three people

Three seats are not optional for the readout, and they should be identified at kickoff so nobody is chasing them in week four.

  • Someone who owns the Azure bill. Cost findings without a budget owner in the room become a document.
  • Someone who owns identity. Security and operational excellence both run through Entra, conditional access, and privileged access, and identity is usually owned by a different team than the one that built the workloads.
  • Someone who gets paged. The person on call knows where the production reality diverges from the diagram. Architects describe intent; on-call describes behavior.

An assessment staffed only with architects produces findings everyone agrees with and nobody funds.

Step 4: Collect the free findings first

Before the engagement starts, action the top Azure Advisor cost recommendations yourself.

Two reasons. You keep the savings, which is the obvious one. And the assessment does not spend billable time re-reporting what the free tool already told you, which is the one that actually changes the value you get.

This feels counterintuitive to suggest before a paid engagement. It is still the right order.

Step 5: Write down what you already believe

Ten minutes, before anyone else looks.

Have your team write down what they think the assessment will find. Where they think the estate is weak. Which workload worries them.

Compare it to the findings in week four. Where the lists agree, you have confirmation and evidence you can take to a budget conversation. Where they diverge is the actual value of an outside look, and you cannot see the divergence if you never wrote down the prior.

Most teams skip this. It is the cheapest step on the list and the one that most changes how the readout lands.

What do you not need to do before a review?

Do not clean up the estate before the assessment.

The instinct is understandable and it defeats the purpose. Tidying the obvious problems in the two weeks before kickoff means the review scores a version of your environment that does not exist the rest of the year. The drift is the finding.

Do not build new documentation either. If the architecture diagram is 18 months stale, hand over the stale one and say so. Documentation drift tells the assessor something real about operational excellence, and rebuilding it under time pressure produces a document that is accurate for exactly one week.

Frequently asked questions

What access does an Azure Well-Architected Review need?

Reader role on the subscriptions in scope, Cost Management read access, and Microsoft Entra ID directory read for the identity portion. Cost Management is the one most often missed because it does not always come with Reader.

How long does preparation take?

For most estates, a few hours spread over the week before kickoff, plus whatever your access-provisioning process takes. Access is the long pole and it is worth starting two weeks out.

Should we fix known problems before the review?

No, with one exception. Action the free Azure Advisor cost recommendations, because you keep those savings and the assessment does not waste time re-reporting them. Beyond that, tidying before an assessment scores an environment that does not exist the rest of the year.

Who needs to attend the readout?

At minimum a budget owner, an identity owner, and someone who is on call for the workloads in scope. Architects alone produce findings nobody funds.

What if our documentation is out of date?

Hand it over as-is and say so. Documentation drift is itself a signal about operational excellence, and rebuilding it under deadline produces something accurate for a week.

What this means for your next 30 days

Start the access request today, even if the engagement is a month out. It is the only step with a queue in front of it.

Then have your team write down what they think you will find. That list is worth more in week four than it looks right now.

Book a scoping call or read the Azure Well-Architected Review offer.

Building something similar?

Talk it through with the engineers who write about it.

Contact us