Reporting periods

The 1-, 7- and 30-day windows, how change indicators are calculated, and which period to use for which question.

Every metric on the dashboard is calculated over a selected window. The window is not a display setting — it changes which answers are in the denominator, and therefore the number itself.

The presets

PeriodBest for
24hChecking today's collection, inspecting individual answers
7dWorking view — enough to smooth daily variance, short enough to notice a change
30dThe default. Trend and reporting; the most stable numbers

A custom range is available from the same picker where you need to line up with a campaign or a release.

How change is calculated

Each metric's delta compares the selected period against the immediately preceding period of the same length. A 7-day view compares this week against last week; a 30-day view compares this month against the previous month.

Two consequences:

  • Changing the period changes the comparison, so a metric can show a rise at 7 days and a fall at 30. Both are correct — they answer different questions.
  • Deltas need history. Until there are two comparable periods of data, values are shown without a change indicator rather than with a misleading one.

For Position, lower is better, so a decrease is presented as an improvement.

Choosing a period

Diagnosing — 24h. You want to read specific answers, not an average.

Working — 7d. Long enough that a single unusual answer does not dominate, short enough to notice a real change within a week.

Reporting — 30d, which is also what the dashboard opens on. Use it for anything shown to someone who will not see the caveats: it is the window least likely to move for reasons unrelated to your work.

Pitfalls

  • Do not compare a 7-day number with a 30-day number. Different windows, different denominators.
  • Short periods plus small prompt sets are volatile. Ten prompts over one day is a handful of answers; treat it as an example, not a measurement.
  • A period that predates a prompt-set change is not comparable. Adding or removing prompts changes what is being measured, so a like-for-like comparison across that boundary does not exist.
  • Empty periods are possible. A period before your account collected anything shows no data — widen the window.

Updated 2026-08-03