How do I report attribution when jobs close weeks after the click?
Report by cohort rather than by calendar month. Group leads by the month they were created, then let revenue accumulate against that cohort as jobs close. A calendar-month report compares this month's spend against revenue that mostly came from last month's leads, which makes growing months look inefficient and shrinking months look unusually profitable.
The arithmetic problem with calendar months
Put spend and revenue in the same monthly row and you have quietly assumed the two arrive together. In repair work that is roughly true. In replacement, install, remodel and any financed job, it is badly false.
The distortion has a direction. When you increase spend, the new leads land this month and their revenue lands next month, so the current month's return looks worse than reality. When you cut spend, revenue from prior leads keeps arriving against a smaller bill, so the month looks great. Managed on calendar months alone, a business gets a steady signal to cut.
What a cohort report looks like
Each row is a lead cohort — leads created in a given week or month, with the spend that produced them. The columns are elapsed time: revenue closed within seven days, thirty days, sixty, ninety, and beyond. The table fills in from the left as time passes.
The join required is not exotic. You need the lead creation timestamp, the source, and every revenue event tied back to that lead's eventual customer and jobs. That tie is the whole engineering job, and it is what a field service integration is for.
One practical warning about building it: a cohort table only works if a lead can be traced to every job that eventually came from it, including the second and third job. If your join stops at the first booking, the table will systematically understate longer cycles and you will conclude that patient channels do not pay.
The maturation curve is the real prize
After a few quarters of cohort data you know what share of a cohort's eventual revenue has typically arrived by day thirty. That single fact lets you evaluate an immature cohort instead of waiting a quarter to find out. A two-week-old cohort with a known curve becomes a forecast with error bars instead of a mystery.
It also tells you when to stop being patient. If a channel's cohort is far behind the curve at the point where most cohorts have matured, waiting longer will not save it.
Segment the curve before you trust it
Emergency repair and planned replacement have different curves in the same company, and averaging them produces a curve that describes neither. Split by job type at minimum, and by financed versus unfinanced if that is a meaningful share of your work. We report these separately inside revenue intelligence for exactly that reason.
Topics: cohorts · sales cycle · lag · reporting
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.