Health report
The Health report in Partner Center lets you get data related to the performance and quality of your app, including crashes and unresponsive events. You can view this data in Partner Center, or download the report to view offline. Where applicable, you can view stack traces and/or CAB files for further debugging.
Alternatively, you can programmatically retrieve the data in this report by using the Microsoft Store analytics REST API.
Crash failure categorization changes for Apps & Games publishers
To enhance the accuracy and quality of failure tracking there are changes planned for Crash Failure categorizations. The changes mentioned will apply to health dataset insights starting on October 15, 2024.
Note
Insights for October 15, 2024 will begin appearing on October 15 or 16, 2024. For insights earlier than October 15, 2024, there will be no alterations.
Because the health datasets display insights from the past 30 days, there will be a transition period between October 15 and November 15, 2024, during which the old and new data buckets will coexist.
These changes would impact Health data set available within Downloads Hub, Public APIs, Partner Center UI & Programmatic APIs for access to Download Hub datasets.
The following changes apply to the Categories that precede the Failure names:
| Categories | Changes |
|---|---|
| - INVALID_POINTER_READ - INVALID_POINTER_WRITE - INVALID_POINTER_EXECUTE - BAD_INSTRUCTION_PTR - NULL_INSTRUCTION_PTR - SOFTWARE_NX_FAULT - ACCESS_VIOLATION_ACCESS_VIOLATION_ |
All the categories mentioned on the left shall be grouped under new category ACCESS_VIOLATION. |
| - ACCESS_VIOLATION_FAIL_FAST | Failures starting with the category is renamed to FAIL_FAST. |
| - STACK_BUFFER_OVERRUN_STACK_BUFFER_OVERRUN | Renamed to STACK_BUFFER_OVERRUN |
| - MISSING_GSFRAME_ - GS_FALSE_POSITIVE_ - GS_POSITIVE_ - GS_NOT_UP2DATE_ - GS_UP2DATE_NOT_KNOWN_ - GS_UNKNOWN_ - GS_COOKIE_INIT_ - GS_SUSPECT_ - ZERO_CANARY_WITH_SE - MINGLY_INTACT_BP_ - NO_MODULE_ - CODE_ - PROBABLY_NOT_USING_GS_ - STACK_COOKIE_CHECK_FAILURE - GLOBAL_COOKIE_CORRUPT_ - INCORRECT_STACK_ - COOKIES_MATCH_ - RUNNING_ON_STACK_ - STRING_DEREFERENCE_ |
The mentioned categories are removed from the failure names. |
Apply filters
Near the top of the page, you can select the time period for which you want to show data. The default selection is 72H (72 hours), but you can choose 30D instead to show data over the last 30 days. Data is shown in your local time zone for the 72H view and in UTC for the 30D view.
You can also expand Filters to filter all of the data on this page by package version, market, and/or device type.
- Package version: The default setting is All. If your app includes more than one package, you can choose a specific one here.
- Market: The default filter is All markets, but you can limit the data to one or more markets.
- Device type: The default setting is All, but you can choose to show data for only one specific device type. The Other category includes devices where the make/model is recognized but we aren't able to include it into one of the predefined categories shown in this filter. For these devices, the device model can be viewed in the Failure log section of the Failure details report.
- OS version: The default is All OS versions, but you can choose a specific OS version.
- OS release version: The default is All OS release versions, but you can choose a specific release version of the selected OS version.
- Sandbox: The default is Retail, but for products that use multiple development sandboxes (such as games that integrate with Xbox Live), you can choose a specific one here. (If your product doesn't use sandboxes, this filter will show only Retail and won't be applicable.)
- Architecture: The default is All architectures, but you can choose a specific system architecture type. This filter is only available when 30D is selected.
- PRAID: The default setting is All, but if you defined multiple package relative app IDs (PRAIDs) when creating your app package, you can choose to show only data related to one PRAID. This filter won't appear if you haven't defined multiple PRAIDs.
The info in all of the charts listed below will reflect the date range and any filters you've selected. Some sections also allow you to apply additional filters.
Failure hits
The Failure hits chart shows the number of daily crashes and events that customers experienced when using your app during the selected period of time. Each type of event that your app experienced is tracked separately: crashes, hangs, JavaScript exceptions, and memory failures.
When the 30D time period is selected, you might see circle markers. These represent a significant increase or decrease in a given value that we think you'll want to know about. The date on which the circle appears represents the end of the week in which we detected a significant increase or decrease compared to the week before that. To see more details about what's changed, hover over the circle.
Tip
You can view more insights related to significant changes over the last 30 days in the Insights report.
Failure hits by market
The Failure hits by market chart shows the total number of crashes and events over the selected period of time by market.
You can view this data in a Table form. Table form will show five markets at a time, sorted either alphabetically or by highest/lowest number of user sessions. You can also download the data to view info for all markets together.
Package version
The Package version chart shows the total number of crashes and events over the selected period of time by package version. By default, we show you the package version that had the most hits on top and continue downward from there. You can reverse this order by toggling the arrow in the Hits column of this chart.
Failures
The Failures chart shows the total number of crashes and events over the selected period of time by failure name. Each failure name is made up of four parts: one or more problem classes, an exception/bug check code, the name of the image/driver where the failure occurred, and the associated function name. By default, we show you the failure that had the most hits on top and continue downward from there. You can reverse this order by toggling the arrow in the Hits column of this chart. For each failure, we also show its percentage of the total number of failures.
Tip
At times, you might see an entry for Unknown in this section. This occurs when despite our best efforts, we're unable to collect full details for one or more failures, which will all be grouped together under Unknown. Most often, this occurs because of storage constraints, but it can also be a result of a device's privacy settings, network connection issues, partial/bad crash dumps, and other factors.
If you see !unknown as part of a failure name, this means that symbols weren't present, so we couldn't identify the failure name. Be sure to include symbols in your package to get accurate failure analysis. See Configure an app package. In contrast, failure names that include !unknown_error_in_ and !unknown_function mean that we weren't able to gather complete details for various other reasons.
To display the Failure details report for a particular failure, select the failure name. If you have included symbol files, the Failure details report includes the number of failure hits over the last month, and a failure log that lists occurrence details (date, package version, device type, device model, OS build) and a link to the stack trace and/or CAB file, if available.
Tip
CAB files will only be available when the failure occurred on a computer using a Windows Insider build, so not all failures will include the CAB download option. To show only failures that have CAB files, select Failures with downloads in the section filter. You can also click the Links header in the Failure log to sort the results so that failures which include CAB files appear at the top of the list.
On the Failure details page, you'll also see the Stack prevalence chart, which shows the top stacks that contributed to the failure, ordered by percentage, and the Device configuration (30D) chart, which provides details about the configuration of devices that experienced the failure.
Crash-free sessions and devices (30D)
The Crash-free sessions and devices chart shows the percent of devices or user sessions that didn't experience a crash in the past 30 days. This info helps you understand how broadly your crashes are affecting your users. For example, an app could have 10,000 crashes in one day. If 90% of your devices are affected, then you would probably classify that as critical and act to fix it right away. However, if that only represents 5% of devices using your app, the priority might be lower.
This chart has two tabs:
- Crash-free devices: Shows the percentage of unique devices that didn't experience a failure on each day (during the past 30 days).
- Crash-free sessions: Shows the percentage of unique user sessions that didn't experience a failure on each day (during the past 30 days).
Crash Rate
A Crash rate is defined as the sum of Crashes occurred for a Game divided by the sum of Activations happened, for a specific timeframe, for all the Users playing the Game. Any time a Game is opened/freshly started/launched for play is counted as one activation. The timeline of an activation and a crash can differ. It may happen that the activation has taken place today, but the crash occurred on the following day. The crash rate is counted only within the mentioned timeframe. It doesn't provide information about - what activation led to which Crash.