How a closed testing campaign runs

Your Google Play account stays yours. You create the closed testing track, upload the build and manage the release. We never ask for access to it, which also means you are never locked in to us in order to ship.

What setup records is the frame everything else is measured against:

• The app and the build under test, so a finding is attributable to a specific release
• The track you will use, and therefore which tester limit applies
• The testing window, with a start date and an end date you have committed to
• The tester count you are aiming for, taken from the current Google requirement for your account type rather than from memory

Most campaigns that fail do not fail because testers could not be found. They fail because the count was wrong: twelve invitations accepted in conversation is not twelve testers opted in.

• Invited: the address is on the roster and the invite has gone out
• Opted in: the tester has accepted with the Google account that was invited
• Active: they have opened the build recently enough to count as participating
• Reporting: they have raised at least one finding against the current build

The count you report is the second state. The count you act on is the third. Keeping them apart is what stops a campaign looking healthy while it quietly is not.

A campaign is healthy when the opted-in tester count stays at or above the threshold you need, continuously, for at least as long as the window requires. Both halves matter: a count that dips in the middle is a different problem from a count that never reached the threshold, and they need different responses.

• Opted-in testers on each day of the window, recorded as it ran rather than reconstructed afterwards
• Days elapsed against the window, so you know how much margin remains
• Any day the count fell short, kept in the record rather than smoothed over
• The date each tester opted in, which is what makes a continuous period provable

When the window closes, the summary should already contain what an application needs:

• The testing window, with exact start and end dates
• Testers opted in, and how many were invited, kept as separate figures
• The number of continuous days the count stayed at or above the threshold
• Any interruption, its date, and what caused it
• The builds tested, and the findings raised against each one

If a figure in your application would not be traceable to a date in this summary, that is a sign the campaign was not recorded closely enough while it was running. A record that is only accurate at its best moment is not evidence of what happened.