Field note
Choosing a fair sample window for in-app search logs
2 September 2025
Why seven days can mislead, and when ninety days is more weight than you need for a first behaviour report.
Search behaviour moves with promotions, seasonal demand, and release cycles. A week that includes a major campaign can look nothing like the following quiet fortnight.
For a first audit we usually prefer thirty to sixty days of anonymised queries. That span is long enough to see recurring intent and short enough that your team can still remember what shipped during the window.
Ninety days helps when traffic is thin or when you need to compare two distinct seasons. It rarely helps on day one if nobody has cleaned the export format yet.
Always note product releases and marketing pushes inside the window. Without that context, a spike in a brand-new phrase can look like an unmet need when it is simply a campaign echo.
Share the window choice with stakeholders before analysis begins. Agreement on the sample is agreement on what the report is allowed to claim.