Definition
A measurement window is the stretch of time a release is measured over. It is set in the measurement plan before launch, next to the audience, the target and the formula. For a team invitations release it might be "first 14 days after launch". The result is read once the window ends and its data is available, and it is compared with the target set before the release shipped.
Every input to the formula is counted inside the same window. If the formula is activated teammates divided by invitations sent, both counts come from the same first 14 days. Counting a month of activations against two weeks of invitations gives a rate that describes neither period.
Why it matters
Without a window agreed in advance, a result can be read on whichever day the number looks best. A launch judged on day 9 because the chart peaked there has not been compared with its plan. Fixing the window before launch, together with the target, gives Growth and engineering the same frame for reading the result.
A window is only as complete as the events recorded inside it. If a required event is missing when the window opens, those days are never collected, and analytics is not retroactive. An event renamed partway through the window splits its count the same way it splits a funnel step. The Renamed event and Removed event entries cover both failures.
How Skene uses it
Skene drafts the measurement window with the rest of the measurement plan, and your team reviews it. Before launch, Skene checks whether the plan's required events are tracked. When the window ends and data is available, Skene compares the observed result with the target and shows where each input came from.
A result above or below the target is evidence for your team's next decision. It does not, by itself, prove the release caused the change.