Microsoft AZ-104: What the Blueprint Covers

AZ-104 preparation is strongest when each Azure service is understood through a concrete administrative action and a verifiable result. Its objectives describe the work required to implement, operate, secure and maintain a real Azure environment. A candidate should know what a feature does, but the stronger test is whether they can select and verify it under the constraints of identity, scope, data access, networking and reliability.

The official Microsoft Learn guide for Microsoft AZ-104 Exam identifies skills measured as of April 17, 2026 for the English assessment. This article interprets the current published coverage and domain weights without inventing a new examination structure or suggesting that a single practice set predicts a pass. It is a source-focused blueprint explanation, distinct from the companion article that follows all five objective groups through one deployment lifecycle.

Read the published weights correctly

Microsoft’s skills at a glance list five groups. These are percentage ranges rather than exact question counts. They establish the approximate emphasis of skill areas; they do not guarantee that every individual question belongs to one group in isolation or that a candidate should divide study time using a precise arithmetic formula.

AZ-104 skill area Published range Underlying administrator question
Manage Azure identities and governance 20–25% Who may manage resources, at which scope and under which policy?
Implement and manage storage 15–20% How should data be stored, protected, accessed and maintained?
Deploy and manage Azure compute resources 20–25% How are workloads deployed, configured and operated?
Implement and manage virtual networking 15–20% How do workloads communicate through intended and restricted paths?
Monitor and maintain Azure resources 10–15% What evidence reveals operational state, and how does recovery work?

The ranges are not an invitation to disregard the smallest domain. A candidate who cannot interpret monitoring or restoration evidence may struggle with compute, storage and network scenarios even if the related facts have been memorized. Similarly, confused identity or scope reasoning can undermine almost any Azure deployment question.

Use the ranges to detect a severe coverage imbalance. Use diagnostic performance, prerequisite gaps and technical evidence to determine what requires more practice. A high weight is a planning clue, not proof that a candidate’s weakest area should automatically be allocated the most time.

Identity and governance: understand who can do what

The first group includes Microsoft Entra users and groups, role-based access control and governance of Azure resources. The key conceptual distinction is between authentication, management authorization and configuration policy. An Azure role can grant the ability to request a deployment; Azure Policy may still deny a resource property that violates an organization’s rules.

Resource scope matters. An assignment at a resource group may permit a different set of actions from an assignment at a subscription or management group. An administrator should understand inheritance and where exceptions must be reviewed, without assuming that broader privileges are an acceptable troubleshooting shortcut.

Know how the Azure Resource Manager hierarchy relates to deployments, identities and policies. A useful scenario asks why an authorized operator’s deployment was rejected. The evidence might identify a policy condition rather than a missing role. Another scenario might involve access to application data, where the relevant permissions differ from those used to manage the storage account.

Beyond vocabulary, be able to articulate one permitted and one prohibited operation for an identity at a stated scope. If both are possible despite the intended restriction, the issue is not solved by merely making sign-in more secure.

Storage: evaluate the data model and its operating constraints

The storage domain covers creating and managing storage resources, configuring access and protection, and working with relevant services such as Azure Blob Storage and Azure Files. The administrator should connect the storage type to the application’s actual pattern: file access, object storage, durability, security, retention, performance and cost.

Data-plane access deserves careful treatment. A management role is not always equivalent to permission to read a particular blob. Shared access mechanisms have scopes and expiration behavior that must be controlled. Network restrictions, private endpoints and DNS can affect whether an otherwise authorized application reaches the storage service.

Use Azure storage examples that distinguish redundancy from historical recovery. Data may be available in more than one location yet still need a backup or retention option after unintended modification. The right answer follows the failure model and recovery requirements, not the redundancy feature with the most reassuring name.

Practice with a synthetic object and a narrowly permitted reader. Verify both a successful legitimate operation and a rejected unauthorized one. This provides better evidence of understanding than being able to list every storage option without knowing its access and recovery implications.

Compute: recognize hosting and operational responsibility

Deploy and manage Azure compute resources includes VM provisioning and operation, hosted applications and container-related options represented in the Microsoft study guide. Candidates should understand deployment settings, runtime responsibilities, scaling, configuration and the consequences of resource changes. The administrator is not choosing a feature in isolation; the workload must operate in a real resource and network context.

A running VM may host a stopped service. A successfully provisioned App Service may fail because the application has the wrong settings or cannot access a dependency. A container may deploy yet not become ready. Distinguish provisioning status from workload health and from the business operation customers need.

Compare hosting alternatives by the responsibility they place on the organization. A VM provides guest-level control while requiring guest maintenance. Managed platforms may reduce some of that work but impose their own configuration and networking requirements. The exam may ask for operationally appropriate configuration rather than a grand solution-architecture debate.

Explain how you would test an intended deployment: confirm the expected version and settings, check the service endpoint, inspect relevant failures and preserve a rollback method. Deployment as code can improve reproducibility, but a template’s success alone cannot prove application behavior.

Networking: trace name, route and security before guessing

The virtual-networking domain includes address spaces, virtual networks and subnetting, connectivity and security controls relevant to Azure administration. Effective routing, name resolution, network security rules and private access can each determine whether a source reaches a destination.

An application cannot read a storage object. It may lack the data-plane role, resolve the wrong endpoint, lack a valid route or encounter filtering. These are different failures. A useful candidate can state what evidence is needed to separate them instead of opening a broad inbound rule or adding unnecessary permissions.

Azure virtual networks provide the underlying model for source and destination paths. Network security groups affect permitted traffic but do not decide whether an authenticated application user should view a particular business record. A private endpoint changes the network path for a supported Azure service, yet still requires correct DNS and authorization conditions.

AZ-104 tests administrator operations rather than the full depth of specialist network engineering. Candidates should be able to implement and diagnose the relevant Azure networking features and know where a more advanced hybrid design may require specialized expertise.

Monitoring and maintenance: observe, respond and restore

The final group includes monitoring and maintaining Azure resources, interpreting health and diagnostics, configuring useful alerts and understanding relevant recovery operations. It receives the smallest published weight range, but every other domain benefits from the ability to distinguish expected state from evidence of actual behavior.

Resource metrics, platform activity records and application logs serve different purposes. A CPU metric can indicate load; an Activity Log event may identify a management change; an application trace may show a failed user operation. Azure Monitor supports collecting and using these observations, but the correct signal depends on what an operator is expected to detect or decide.

Maintenance also includes recovery thinking. Scaling can help capacity, redundancy can protect against some infrastructure failures, and backups preserve recovery points when properly configured and tested. None automatically proves that a complete application can return to service with acceptable downtime and data loss.

For a meaningful test, identify a small synthetic resource failure, observe the expected signal and demonstrate a controlled correction. Then verify that the permitted workload works and that intended security boundaries remain intact. A green dashboard after a change is insufficient if the underlying application still fails or has gained unintended access.

Interpret objective verbs as expectations of competence

The guide’s task language matters. To configure a resource, a candidate must understand the required settings and relevant limits. To manage it, they must understand how it changes under ordinary operations. To monitor it, they must know what evidence represents state and failure. To implement security, they must specify where the control is enforced and how to test both allowed and disallowed outcomes.

A learning resource that defines every term without showing those decisions may be factually accurate but insufficient for the exam. Conversely, memorizing portal steps without understanding why a setting matters may fail as soon as the scenario changes. Mix explanation, interpretation and controlled practice rather than allowing familiar question wording to stand in for competence.

The guide’s examples illustrate skills; related topics may also appear. Microsoft’s documentation notes that most questions concern generally available features and some widely used preview features may be included. Avoid making unsupported claims that every item must match one named product release or exactly one sample configuration.

Control the version and finish with broad coverage

Microsoft changes its exams to reflect role requirements. The current English guide identifies April 17, 2026 as the operative skills date. The document also describes its change log against the previous version, with largely limited changes in several skill groups. A candidate should preserve that version distinction when using third-party notes created under an older guide.

Localized examinations may be updated after the English version, and specific availability or testing rules should be checked on the official exam listing. Do not infer a new exam code or retirement simply because a skills guide received an update. Likewise, do not treat an old AZ-103 article as a current examination blueprint.

A good final check samples all five groups with unfamiliar decisions. Can you explain a denied deployment, a blocked storage read, an unhealthy application, an unexpected network path and a failed restoration using the evidence actually provided? The AZ-104 blueprint is useful when it turns those individual cases into one picture of competent Azure administration: authorized deployment, controlled access, working services, observable state and safe recovery.