Finding Why a User's Sign-In Failed in the Microsoft Entra Sign-In Logs
A user's Entra sign-in failed? Filter the sign-in logs by user and Failure, read the error code and failure reason, and know where codes 50126 and 53003 lead.

Advertising disclosure: this article includes partner links. KBY may earn a commission if you sign up or buy through one, at no extra cost to you. How we make money.
Before you begin
- A lab or test Microsoft Entra tenant.
- An account with the Reports Reader role.
- A lab test user whose password you may use for the exercise.
To find out why a sign-in failed, open the Microsoft Entra admin center as at least a Reports Reader, browse to Entra ID > Monitoring & health > Sign-in logs, filter on the user and set Status to Failure. Select the row. The Basic info tab shows the Sign-in error code and the Failure reason, and the failure reason or Additional Details often says what to do next. Error 50126, for example, means the user typed the wrong username or password.
The constraint that catches people is time. With Microsoft Entra ID Free the sign-in log keeps seven days; with P1 or P2 it keeps 30. A failure the user reports two weeks later on a Free tenant is already gone, unless the logs were sent somewhere else. Entries are generated by the system and can’t be changed or deleted, which is why they are worth trusting. This guide explains what a sign-in row records, walks through one failure from start to finish, and has you produce and read a failure of your own on a lab account.
#What you will learn from one failed sign-in
- Describe a sign-in as who, how and what, and say which part a given failure belongs to.
- Find a single failed sign-in by user, status and time, and read its error code, failure reason and correlation ID.
- Decide where to look next for the common codes, including Conditional Access and MFA failures.
- Read the same record from Microsoft Graph PowerShell when you need it in a script.
- Know when to stop and hand the case on, and what to hand over.
#What you need to read the sign-in logs
- A lab or test Microsoft Entra tenant. Sign-in logs are available on every edition, Free included.
- An account with the Reports Reader role, the least privileged role Microsoft lists for reading the activity logs. Microsoft also lists Global Reader among the roles that can read sign-ins through Graph.
- A test user in that tenant whose password you know and are allowed to use for the exercise. Never use a real colleague’s account.
- For the PowerShell section only: the Microsoft Graph PowerShell SDK, a P1 or P2 tenant (Microsoft documents Graph access to the activity logs for P1 and P2 tenants) and consent for the
AuditLog.Read.Allpermission.
#How a sign-in becomes a log row
Every time someone or something asks Microsoft Entra ID for a token, Entra decides whether to issue it and records the attempt. Microsoft describes each attempt with three parts. Who is the identity signing in: the user, with their username, user ID and the sign-in identifier they typed. How is the client used: a browser, a desktop or mobile app, or a legacy protocol such as IMAP or SMTP. What is the resource the token is for, such as a Microsoft 365 service or your own enterprise application. Most failures belong to one of the three. A wrong password is a who problem, a blocked legacy client is a how problem, and a user who isn’t assigned to an application is a what problem.
The admin center splits the log into four tabs. Interactive user sign-ins are the ones where a person provides a factor: a password, an MFA response, a biometric. Non-interactive user sign-ins are made by a client on the user’s behalf with tokens it already holds, so nobody typed anything. Service principal sign-ins and Managed identity sign-ins involve no user at all. When a user says they typed their password and it failed, start on the interactive tab. When an app keeps failing in the background without prompting, the non-interactive tab is the likelier place.
Each row also carries identifiers that tie it to other records. The Correlation ID groups the requests that belong to one sign-in session; it comes from the client, so Microsoft doesn’t guarantee its accuracy. The Request ID corresponds to an issued token. If you ever escalate a failure to a colleague or to Microsoft, these two values are what let them find the exact attempt you are looking at.
#The fields that explain a failure
Open a failed row and the Activity Details pane appears with several tabs. Basic info holds most of what you need: the user, the application, the resource, the status, the Sign-in error code, the Failure reason, the identifiers above and a link to Launch the Sign-in diagnostic. Location and Device info show the IP address, browser and operating system, and whether the device was compliant, managed or hybrid joined. Authentication Details lists each step of authentication in order, such as the password and then the MFA method, and whether each one succeeded. Conditional Access lists every policy that could apply, with a result of Success, Failure, Not Applied or Disabled.
The failure reason is a sentence written for people; the error code is a number written for lookups. Read the reason first. If it doesn’t tell you enough, put the code into Microsoft’s error code lookup tool or search for it in the sign-in error codes reference, where codes appear with an AADSTS prefix. The table lists codes that come up often in user sign-ins and where each one points you.
| Code | What Microsoft says it means | Where to look next |
|---|---|---|
| 50126 | Invalid username or password. Microsoft expects some of these in every log because people mistype. | Ask the user to try again carefully. Repeated failures from places the user doesn’t recognise are a security question, not a typing one. |
| 50057 | The user account is disabled. | Find out who disabled it and why before anyone enables it again. |
| 50053 | The account is locked after too many failed attempts, or the sign-in came from an IP address with malicious activity. | The failure reason tells you which of the two it was. |
| 53003 | Blocked by Conditional Access policies. | The Conditional Access tab, to see which policy returned Failure. |
| 500121 and 50074 | The user didn’t complete or didn’t pass the MFA prompt. | Authentication Details. 500121 often means the user hasn’t finished setting up MFA. |
| 50055 | The password has expired. | The user is offered a reset, or an admin resets it. |
Some codes aren’t failures the user would notice. Microsoft describes 50058 as common and expected when a user isn’t signed in yet; there is a separate guide to what AADSTS50058 means for silent sign-in. A code like 50105, where the user isn’t assigned to the application, is a what problem fixed in the enterprise application rather than on the user; see fixing AADSTS50105 for an unassigned user.

#Deciding where to look next
Once you have the code, the next step depends on which part of the sign-in it points at. The flowchart below follows the codes in the table to the tab or person that can answer the question, and ends with the escalation route for anything you can’t place.
Rendering diagram...
Notice that none of the branches start with changing something. Every one of them is another read: a tab, a failure reason, a conversation with the account owner. The fix comes after you know which branch you are on.
#An example from the service desk
Suppose a user, Sam, reports at 09:40 that they can’t open the company’s expenses app from home, although it worked in the office the day before. You open the sign-in logs, add a User filter for Sam’s user principal name, set Status to Failure and set the date range to the last 24 hours. One row matches, at 09:31, for the expenses application.
On Basic info the sign-in error code is 53003 and the failure reason says access was blocked by Conditional Access policies. That points away from the password: the code is about a policy decision, not about the credentials Sam typed. On the Conditional Access tab, one policy shows Failure and the others show Not Applied. The Location tab shows an IP address that isn’t one of the office’s. The conclusion the evidence supports is narrow: a Conditional Access policy that applies to this app and this sign-in did not allow it, and the difference from yesterday is the network. Whether Sam should be allowed in from home is a policy decision for whoever owns that policy, not something to change at the service desk. You pass on the policy name, the time, the correlation ID and the request ID.
The 09:31 on that row is in your time zone. Microsoft localises the admin center’s timestamps to the person viewing them, not the person who signed in. If Sam is in a different time zone, say so in the hand-over.
#Practice on a lab account with one wrong password
The outcome you want is one failure you understand completely, found and read by you. Nothing in the tenant’s configuration changes. The only lasting effect is new entries in the sign-in log, which can’t be deleted.
- Note the current time. Open a private browser window, so your admin session isn’t reused.
- Go to
https://mysignins.microsoft.comand sign in as the lab test user, typing the password wrong once. Don’t try again with a wrong password: repeated failures are how an account gets locked (50053), and that is not the lesson here. - Sign in again as the same user with the right password, completing MFA if the tenant asks for it, then close the private window.
- In your admin session, open Entra ID > Monitoring & health > Sign-in logs. Stay on Interactive user sign-ins, add a User filter for the test user and set the date range to the last 24 hours.
- Find two rows near the time you noted: one Failure, one Success. Open the failure.
- On Basic info, write down the sign-in error code, the failure reason and the correlation ID. Then open Authentication Details and see which step it records. Microsoft notes this tab can show incomplete data until the log is fully aggregated, so a missing row there is not a finding on its own.
Expected evidence: the failed row shows error code 50126 and a failure reason along the lines of “Error validating credentials due to invalid username or password”. The second row shows a status of Success. You pass when you can say, from the log alone, that the first attempt failed on the password and the second attempt succeeded.
Stop if the failed row shows anything other than 50126, such as 50053 or 50057: the test account is not in the state you thought, so leave it alone and tell whoever owns the lab tenant. If no rows appear, check the user filter, the date range and that you are on the interactive tab before assuming anything is broken. There is nothing to roll back in the tenant. Close the private window; if you used PowerShell, run Disconnect-MgGraph.
#Reading the same row from PowerShell
The admin center is the right tool for one user and one incident. When you need the same answer in a script, Microsoft Graph returns the same records as signIn objects. The status property holds errorCode, failureReason and additionalDetails. Sign in with the read permission the sign-ins API needs:
Connect-MgGraph -Scopes "AuditLog.Read.All"The first time, you may be asked to consent to AuditLog.Read.All. Microsoft documents that consenting to the permissions for reading logs through Graph needs a Privileged Role Administrator. If the prompt asks for consent you can’t give, stop and ask; don’t look for another account that can. When sign-in works, PowerShell prints a welcome message. Then list the test user’s sign-ins since the start of the day, newest first. Replace the user principal name with your test user’s and the date with today’s:
1Get-MgAuditLogSignIn -Filter "userPrincipalName eq 'lab.user@contoso.onmicrosoft.com' and createdDateTime ge 2026-11-22T00:00:00Z" -Top 10 |
2 Select-Object CreatedDateTime, AppDisplayName, CorrelationId,
3 @{n='ErrorCode';e={$_.Status.ErrorCode}},
4 @{n='FailureReason';e={$_.Status.FailureReason}}Both filters are ones the API documents: userPrincipalName supports eq, and createdDateTime supports ge and le. Microsoft recommends filtering on a time range so the request doesn’t time out. You should see the same failure with error code 50126 and the same correlation ID you wrote down in the portal. Here CreatedDateTime is in UTC, not your local time, so the hour may not match the portal. When you are done, run:
Disconnect-MgGraph
#How to check you read the right row
A sign-in log is easy to misread because many rows look alike. Before you report a cause, check three things. The user principal name on the row is the user who reported the problem; Graph always stores it in lower case, and guest users appear with their home address rather than the #EXT# form. The application and resource are the ones the user was trying to reach, not another app signing in around the same time. The time matches what the user described once you allow for your own time zone in the portal and UTC in Graph.
Then check that the conclusion matches the field you took it from. A 53003 with one policy showing Failure supports “Conditional Access blocked this sign-in”. It does not, on its own, support “the policy is wrong”. A 50126 supports “the password was wrong for this attempt”. It does not tell you whether the user or someone else typed it.
#Where it goes wrong when reading sign-in logs
- Looking after the log has expired. A Free tenant keeps seven days. Upgrading later does not bring back older data.
- Searching only the interactive tab for a failure that happened in the background. Token refreshes by apps appear under non-interactive sign-ins.
- Reading a Conditional Access status of Success as “a policy granted access”. Microsoft documents that Success can also mean a policy was evaluated without applying.
- Expecting the Conditional Access details through Graph with only Reports Reader. The applied policies are only returned to roles that can read Conditional Access data, such as Global Reader or Security Reader.
- Treating the location as fact. Microsoft calls IP-to-location conversion best effort; VPNs and mobile networks often hand out addresses far from where the device really is.
- Expecting one MFA row per prompt. The admin center shows a multi-step MFA sign-in as one line item; the same event in Azure Monitor appears as several, all with the same correlation ID.
#What to remember about failure codes
- Filter by user, status and time, then read the failure reason before the code.
- Place the failure in who, how or what before deciding who can fix it.
- The correlation ID and request ID are what make an escalation findable.
- Seven days on Free, 30 on P1 or P2: ask for the time of the failure first.
- Reading the log changes nothing, and the log itself can’t be edited.
#Before you look at a real user’s sign-ins
Sign-in logs hold personal data: names, IP addresses, locations and devices. In production, ask for the Reports Reader role rather than anything broader, look only at the users and times a ticket gives you a reason to look at, and don’t paste raw rows into chat channels. Users can see their own sign-ins at https://mysignins.microsoft.com, which is often the quickest way for them to confirm an attempt was theirs.
Know where your part ends. If the cause is a disabled account, a Conditional Access policy or a lockout from a suspicious IP address, the next step belongs to the account owner, the policy owner or the security team. Hand over the user principal name, the time with its time zone, the application, the error code, the failure reason, the correlation ID and the request ID. If the failure reason and the error code reference don’t explain it, launch the Sign-in diagnostic from the row; if that doesn’t either, a support request with those same values is the next step. If the user was deleted rather than disabled, restoring a deleted user within the 30-day window covers what can still be recovered.
About this page
Credited to
Dates
Published
Sources
Corrections
No corrections are listed for this page. Some review sweeps corrected pages without naming each one; those are in the corrections log.
Related articles
Identity and Access Management
Entra ID AADSTS50058: Silent Sign-In Request Sent With No Session
Seeing AADSTS50058, silent sign-in request sent but no user signed in? Learn how session cookies and prompt=none work, and test an interactive fallback.
ReadEnterprise IT Management
Part 3: Create a Windows Autopilot Deployment Profile and Dynamic Device Group in Intune
Build a dynamic device group on the AP-Pilot group tag, assign a user-driven Microsoft Entra join profile with a standard user account and a name template, and wait for Assigned before deploying.
ReadArtificial Intelligence
GitHub Lets Enterprise Admins Force Copilot Customisations Through Plugins, and Streams Agent Activity to Microsoft Agent 365
GitHub added a plugin-only setting for Copilot skills, agents, hooks, MCP servers and extensions, audit streaming into Microsoft Agent 365, Entra sign-in for Copilot CLI and team defaults. Secret scanning is available but off. What to decide first.
ReadDiscover more
Graduate Learning
- TutorialCreating a Dynamic Membership Group in Microsoft Entra ID and Validating Its Rule
- TutorialRestoring a Deleted User in Microsoft Entra ID Within the 30-Day Window
- TutorialEntra ID AADSTS50058: Silent Sign-In Request Sent With No Session
- TutorialEntra ID AADSTS7000215: Invalid Client Secret Provided
Ops Playbook
Lexicon Definitions
Was this useful?
Reports go to the site owner and are used to correct pages. Corrections are listed at /corrections.
Build practical engineering skills.
Receive new lessons, learning paths, practical exercises and early-career guidance.
Comments
Add a thoughtful note on Finding Why a User's Sign-In Failed in the Microsoft Entra Sign-In Logs. Comments are checked for spam and held for moderation before appearing.