Microsoft AZ-900: Readiness by Exam Domain

Microsoft AZ-900 readiness is not an estimate of how likely an individual is to pass. It is evidence that the learner can apply current cloud concepts, Azure architecture and management controls to unfamiliar situations. The matrix below helps record that evidence under the study guide effective July 20, 2026, rather than rewarding repeated exposure to the same practice answers.

The current outline assigns 25–30% to cloud concepts, 35–40% to Azure architecture and services and 30–35% to management and governance. These ranges are useful for prioritizing gaps but do not reveal exactly how many questions an individual will receive. Domain reviews should produce explanations and corrections, not an invented numerical score.

Establish an evidence scale that is difficult to game

Mark an objective green when you can explain it accurately in a new scenario, identify a tempting wrong answer and justify the choice without notes. Mark it amber when the service name is familiar but the underlying boundary or limitation is unclear. Mark it red when the intended outcome is misunderstood or the same distinction repeatedly causes errors.

For every item, document the business requirement, selected principle and reason an alternative fails. A learner may be comfortable with a portal but have poor understanding of shared responsibility; an experienced systems engineer may know VMs but confuse Azure Policy and RBAC. Separate these gaps rather than treating general experience as uniform readiness.

An objective should not become green merely because a module or video is completed. Transfer the idea to a new industry or technical constraint, and verify that the explanation still holds. This avoids a matrix that measures repetition rather than understanding.

Cloud concepts: deployment, service models and operating benefits

Test public, private and hybrid deployment models separately from IaaS, PaaS and SaaS. Given a company that keeps a datacenter application while hosting a managed web service in Azure, identify both dimensions. Explain who patches the operating system and who protects the records under each service model.

Next, distinguish consumption-based costs, capital versus operating expenditure, and serverless execution. A candidate should explain how variable demand can make elastic capacity attractive while persistent storage or uncontrolled test resources can still raise costs. Cloud should not be described as always cheaper or automatically secure.

Test availability, reliability, scalability, elasticity and predictability using different failures. A website can scale without being able to restore a deleted database, and redundant hardware can still suffer a shared misconfiguration. Azure cloud concepts are understood when the candidate can name the specific mechanism behind each promised benefit.

Architecture: scope, regions and services

Draw both a geographic and an administrative map. Regions, datacenters and availability zones are placement and fault concepts; management groups, subscriptions, resource groups and resources are governance and deployment scopes. Do not confuse a resource group with a physical availability boundary or a subscription with a virtual network.

For compute, use cases should force a choice among VM, container, web app and function. Explain when guest OS control matters and when managed runtime operations are a benefit. For networking, trace the role of VNets, subnets, peering, DNS, public and private endpoints, VPN Gateway and ExpressRoute without assuming a VNet automatically authorizes data access.

For storage, select blob, file or disk according to the workload; then assess tiering and redundancy as separate questions. Recognize AzCopy, Azure Storage Explorer, File Sync, Azure Migrate and Data Box by purpose. An explanation that treats a transfer utility as a redundancy setting remains incomplete.

Identity and security: who can do what?

Distinguish Microsoft Entra ID, Entra Domain Services, sign-in mechanisms and authorization. Compare MFA, passwordless authentication, SSO and Conditional Access with a case in which a user’s sign-in succeeds but access to an Azure resource is still denied. A valid answer identifies the relevant control plane.

Use distinct requirements for external identities, RBAC, Zero Trust, defense in depth and Microsoft Defender for Cloud. A trainee who needs read-only visibility should receive an appropriately scoped role, rather than an unrestricted administrator grant for convenience.

A security case is green only when the candidate connects a named control to the identity, resource and operation at risk. Familiar acronyms are not evidence that an access request or security posture has been understood.

Governance and cost: controls with different purposes

A pricing calculator estimates spending for a hypothetical configuration; Cost Management helps track spending and budgets. Tags can identify ownership but do not enforce who can modify a resource. Azure Policy evaluates or restricts resource configuration, resource locks help prevent certain management-plane changes, and Microsoft Purview serves data governance and compliance purposes.

Create a hypothetical change request: all new resources must use an approved region; a production resource must resist accidental deletion; and costs must be allocated to a business team. These lead to different controls, not one universal governance product.

Explain the Azure portal, Cloud Shell, CLI, PowerShell, Azure Arc, Azure Resource Manager and infrastructure as code at the level of their purposes. A candidate should know which provides an interface, a command tool, management beyond Azure or reproducible resource definitions without writing a complete deployment script.

Monitoring: turn service names into questions

Use one performance issue, one provider service incident and one optimization opportunity. Explain how Azure Monitor, including logs, alerts and Application Insights, differs from Azure Service Health and Advisor. A question about a slow application is not automatically a question about a public Azure outage.

Azure operational monitoring becomes meaningful when the candidate can state what evidence is required, who would receive it and what conclusion it can support. A dashboard screenshot is weak evidence unless the user can explain what the data means.

Add at least one case where the apparently familiar answer is wrong. For example, a rising monthly bill needs cost analysis rather than a service incident notification. A resource recommendation is not the same as an application trace.

Check the whole matrix with fresh scenarios

Select unfamiliar cases spanning all three domains. State the decisive requirement before viewing possible answers. Do not award green to a correct option supported by an incorrect explanation; mark it amber and repair the principle. Domain weights should guide attention, not be converted into a homemade pass score.

Microsoft’s official Practice Assessment and exam sandbox are useful for wording and interface practice. Compare their feedback with your explanations, then use AZ-900 hands-on scenarios where a new observation would resolve an uncertain concept. Repeating one memorized bank does not establish independent understanding.

If the same error occurs after correction, diagnose whether the problem is a wrong mental model or a missed detail. The targeted AZ-900 preparation mistakes help connect recurring issues with a corrective exercise, without launching a second generic study plan.

Before booking, look for sustained new-scenario explanations, no unresolved red core domains and familiarity with the official testing environment. This supports a practical readiness decision but does not guarantee success. Verify actual exam language, identification and scheduling rules with Microsoft rather than older blog claims.