How long should our attribution lookback window be?
Long enough to cover the real gap between first contact and the sale, and no longer. Measure the actual distribution of days from lead creation to booked job in your own system, then set the window near the point where most jobs have closed. Too short and long-cycle work looks unprofitable. Too long and unrelated touches collect credit. Report emergency and planned work separately, because their distributions barely overlap.
Measure the lag before you pick the window
Almost nobody does the one step that settles this: pull every job from the last year, compute the days between lead creation and the booked or completed date, and look at the distribution. It is usually not a bell curve. It is a tall spike in the first day or two and a long thin tail running out for weeks.
The spike is emergency work. The tail is planned work, estimates that needed a decision, and quotes that sat on a kitchen counter. Averaging them produces a number that describes neither.
Conversion lag makes recent periods look worse than they are
If jobs keep closing for three weeks after the lead, then the last three weeks of any report are incomplete by construction. Revenue for those weeks will keep filling in after the report is printed.
Managers who do not know this reliably conclude that performance is declining, because the most recent bar on the chart is always the shortest. The fix is to show maturation explicitly: mark recent periods as incomplete, or report on a lag so that every period shown has fully matured. This is one of the most common causes of an unnecessary budget panic, and it is easy to design out of a dashboard.
Different windows for different questions
- Bidding signal. Wants short and fast. Use a mid-funnel event that occurs within days, such as the booked appointment.
- Channel reporting. Wants the window that covers most closed jobs, applied consistently and stated on the report.
- Customer value analysis. Wants a much longer horizon, because repeat work and referrals arrive months later and belong to a different question entirely.
- Platform limits. Ad platforms cap how far back a click can be credited, so your operational window may exceed what any platform will honor.
Split the report by job type
Once you separate emergency from planned work, most of the argument disappears. Emergency jobs settle within days and a short window is fine. Planned jobs need weeks, and forcing them into the same window makes every channel that produces considered demand look weak.
For contractors running both, the practical setup is one window per job category, documented, with the category assigned from the job type in the operational system rather than guessed from the campaign. Doing that reliably is ordinary field service system integration work, and it changes which campaigns look profitable.
Topics: lookback window · conversion lag · reporting periods · sales cycle
Have a version of this question about your own business?
The useful answer usually depends on which systems you run and how they're connected. That's a conversation, not a blog post.