Microsoft MD-102: Endpoint Operations and Troubleshooting

Microsoft MD-102 is not just an enrollment-and-policy exam. The endpoint administrator has to operate a device service over time: establish identity and ownership, enroll devices, apply configuration, deploy applications, enforce security, evaluate compliance, maintain Windows, monitor health, automate repetitive work, and troubleshoot failures that cross those layers.

As of October 6, 2026, MD-102 is current on the July 24 English blueprint; Microsoft has published an update effective October 27. This article focuses on operating and troubleshooting endpoints, which keeps it distinct from the separate objectives, study-plan, difficulty, lifecycle, and identity/compliance articles already in the cluster.

Start troubleshooting with device state

Before changing policy, establish platform, ownership, join or registration state, enrollment status, primary user, management authority, and last successful check-in. Use one failed configuration scenario and write down the expected device state versus the observed state before touching any profile or assignment.

Many apparent policy failures are really enrollment, identity, scope, licensing, or communication problems.

Device state should be established before any policy is changed. Platform, ownership, join state, enrollment state, primary user, management authority, and last check-in can explain why a policy never arrives. Changing the policy when the device is not in the expected state only adds noise.

Trace enrollment as a dependency chain

The MD-102 enrollment and device lifecycle model helps separate preparation, identity, enrollment, assignment, configuration, maintenance, and retirement. For a device that never becomes manageable, check prerequisites in order: supported platform, identity, enrollment restriction, ownership, license, network access, and service contact.

Troubleshooting is faster when you locate the first failed state transition instead of repeatedly reissuing downstream policy.

Treat assignment as part of configuration

A profile can be perfectly configured and still fail because the intended device or user never enters the assignment scope, is filtered out, or receives a conflicting setting. Compare user-targeted and device-targeted assignments, dynamic-group timing, filters, exclusions, and applicability for the same setting.

The evidence should show both that the policy exists and that the target actually qualifies to receive it.

Assignment troubleshooting should verify that the target qualifies. User or device scope, dynamic-group timing, filters, exclusions, applicability, and conflicts can all prevent a valid setting from reaching the intended endpoint. The admin should prove the assignment path instead of assuming the configuration itself is wrong.

Separate compliance from Conditional Access

MD-102 compliance and Conditional Access connects endpoint state to access but uses two different decision points: Intune evaluates compliance and Entra can consume that result in access policy. When a user is blocked, verify device compliance state and timing first, then inspect Conditional Access evaluation and application authorization separately.

This prevents support teams from changing an Intune compliance policy when the denial actually belongs to Entra or the target application.

Use one final mixed endpoint case that includes a newly provisioned device, an application problem, stale compliance, and an access complaint. Resolve the issues in dependency order and record where ownership moves from endpoint management to identity or security.

Include one post-remediation verification step in every playbook. After a fix, confirm that the intended device state, application status, compliance result, or access path is actually restored rather than assuming the configuration change succeeded.

Diagnose application delivery by context

Application deployment depends on platform, install context, detection rules, dependencies, requirements, assignments, supersedence or update logic, and sometimes user interaction. For a failed deployment, identify whether the app never targeted, could not download, failed installation, installed but was not detected, or later became unhealthy.

Those failure categories point to different evidence and keep application troubleshooting from becoming random repackaging. A useful diagnose application delivery by context summary therefore records why the choice fits the MD-102 endpoint-operations requirement and which assumption would trigger a redesign.

Application failures should be classified by stage: not targeted, unable to download, failed installation, installed but not detected, or later unhealthy. Each stage points to different evidence. That classification is more useful than repeatedly repackaging an application whose real problem is scope or detection logic.

Application troubleshooting should confirm assignment, applicability, download, installation, detection, and post-install health as separate stages. A package that never reached the device is a different problem from one that installed but failed detection. Recording the stage and evidence makes the playbook faster and prevents repeated repackaging when the actual issue is targeting or detection logic.

Use Autopilot evidence instead of assumptions

Windows Autopilot deployment modes differ in who drives provisioning, when device identity is known, which stages must finish, and what the user sees. For a provisioning failure, identify the failing phase, assigned profile, network dependency, enrollment status, required applications, and whether the device should be reset or the policy corrected.

A screenshot of a stuck page is not a diagnosis; the useful artifact is the stage and dependency that failed.

Operate endpoint security as managed state

Encryption, firewall, attack surface reduction, Defender for Endpoint, privilege controls, and security baselines need configuration, telemetry, exception handling, and ownership. Take one endpoint security control and map desired state, enforcement path, reporting signal, exception process, and incident escalation path.

This separates policy drift from active compromise and clarifies when endpoint administration hands off to security operations.

Operational maturity also means knowing when not to remediate automatically. A safe endpoint workflow distinguishes low-risk drift that can be corrected at scale from identity, security, or data-loss symptoms that require investigation first. During preparation, mark which actions are reversible, which can disrupt users, and which could destroy evidence. That decision discipline is useful both for MD-102 scenarios and for real endpoint operations.

Endpoint automation should be used cautiously. Repetitive low-risk drift can be detected and remediated at scale, but identity, security, or data-loss symptoms may require investigation first. The playbook should make that boundary visible so automation does not erase evidence or disrupt users unnecessarily.

Security and endpoint administration need clear handoffs. A configuration drift problem may be remediated automatically, while suspicious endpoint behavior may need to be preserved for investigation before change. The administrator should know when a policy issue becomes a security incident and which evidence must be retained.

Security configuration should be verified as managed state, not assumed from policy assignment. Check whether Defender, firewall, encryption, local privilege, and attack-surface settings are actually present on the intended device population and whether any exception has an owner. That evidence helps distinguish policy-delivery problems from a security incident or a legitimate approved deviation.

Use analytics and automation to manage by exception

Endpoint analytics, reports, remediation, service health, alerts, PowerShell, and Microsoft Graph can turn repeated support work into detectable conditions and controlled corrective actions. Choose a recurring endpoint issue, define the signal that identifies affected devices, automate a safe remediation, and measure whether the population returns to desired state.

The administrator’s leverage comes from reducing repeated manual intervention without hiding failures.

Build a troubleshooting playbook from evidence

A useful troubleshooting playbook starts with the symptom, identifies the owning layer, lists the minimum evidence, and defines the next branch from that evidence. Microsoft 365 Certified: Endpoint Administrator Associate reflects that operating responsibility across enrollment, applications, compliance, security, maintenance, and support.

If the playbook tells you which evidence to inspect before which setting to change, it is useful. If it is only a list of portals, it is not.

Build one end-to-end support case that crosses multiple endpoint layers. Start with a newly provisioned device, introduce an application failure, then make the device noncompliant and finally add a user access complaint. For each symptom, identify the first evidence source you would inspect and the team boundary you would respect. This prevents the common habit of treating every endpoint problem as an Intune profile problem and gives you a repeatable method for finding the first broken state.

Troubleshooting should record the first unexpected state, not just the final fix. If a device is enrolled but never receives a policy, that is a different failure from a policy that arrives but conflicts with another setting. Writing the first wrong state makes later incidents faster because the playbook points to evidence instead of a list of portals.

Verify recovery and close the endpoint lifecycle

Maintenance should include retirement. Old devices, stale primary-user relationships, obsolete applications, and abandoned configuration can remain long after normal operations move on. A complete endpoint lifecycle includes removal and cleanup so historical state does not keep generating risk or support noise.

Support quality improves when device evidence is captured before remediation. Record policy state, compliance result, application status, relevant event or management logs, and recent changes before resetting or re-enrolling a device. That evidence helps distinguish a one-off endpoint failure from a wider assignment, service, or policy problem.

Verify the repaired endpoint state after every remediation so the playbook proves recovery, not just configuration change.

A useful endpoint playbook should record the first unexpected state, the evidence that proves it, the owning control plane, and the safest next action. Include one device-state failure, one application-delivery failure, one compliance/access disagreement, and one security symptom. After remediation, verify that the intended state actually returned and that the fix reached the correct population. This prevents support procedures from becoming lists of resets and sync commands that restore service without explaining or preventing the underlying failure.

  • img