Learn from real use
Keep privacy-safe run records and summary data with the exact action version. Separate “it worked” from “people use it,” “it helped,” and “this release caused the change.”
Start with one run
Measure starts with the newest local run. The first question is simple: what happened, which action version ran, and what record supports that answer?
Summary data comes next when it helps explain the run. A result can lead back to Build or the next Audit without pretending that limited data proves everything.
What you can review
The exact action and version, whether it worked, how long it took, and what started it when known.
Successes, failures, cancellations, and simple error groups without saving private inputs or content.
Privacy-protected Apple summaries with clear dates, app builds, minimum counts, and reporting delays.
Optional, consented links to product results that stay separate from proof that the action merely ran.
Four separate questions
A completed run can show that the action worked. By itself, it cannot show that people keep using it, that they got the intended result, or that a release caused a wider change.
Run records and privacy-safe success totals
RecordedPrivacy-safe summary counts
LimitedApproved product data
OptionalA controlled test and permission to use the data
UnknownWhere the data comes from
Local runs, Apple summary reports, outside tools, and optional business results can describe the same app action. They do not all prove the same thing.
Run records for the exact action version, result, time, and caller when known.
AvailablePrivacy-safe counts and success totals that may need a minimum amount of data and arrive late.
ImportMCP, API, or command-line records from a connection that has been tested.
Source recordedA business or user result from a separate, approved data connection.
Not connectedPrivacy by default
The local record works without a network connection. You control how long it is kept, how it is exported, and when it is deleted. Linking outside business results requires separate consent.
Read the Measure documentationCompare releases carefully
A difference between releases can point to the next question. It does not prove that the release caused the difference without a stronger test and the right consent.
Close the loop
A run record can show a reliability problem, missing data, or an open product question. Send that finding into the next Audit with the action version and supporting record attached.
Plan what to measure
Choose what to record, what to leave private, and what each source can honestly tell you.