WHERE THE COHORT LEAKS

The core of the map is one table: for customers who first ordered in a given month, how many reached each stage.

Why cohorts, not calendar months

A calendar-month report mixes customers at every point in their journey: acquire twice as many in March and April’s second orders rise even if nothing improved. A first-order cohort follows one group from the same starting line. The Second Order covers cohorts in depth; here you need one table.

Give each stage a fixed window, measured from the first order, so cohorts compare fairly: delivered by day 14, first use by day 30, second order by day 120, third order by month 12. A cohort can only report a stage once its window has closed.

A worked example

Say a brand’s March cohort has 5,000 first-time customers. 4,900 are delivered (the rest canceled or were refunded before delivery). 3,000 show the first-use signal by day 30. 1,050 place a second order by day 120, and 600 a third by month 12.

Rank by customers lost. The step to the third order loses 43% of those who reach it, alarming in a rate report, but only 450 people. The steps into first use and the second order lose 1,900 and 1,950. With two stages this close, the next section breaks the tie.

Customers, contribution and the value of a point

Not every lost customer is worth the same. One lost before first use still had two stages to pass before a third order; one lost at the second order had one. A point gained early reaches more customers, but fewer of them go on. In this example, a 5-point gain at the second order is worth about $123,000 a year and the same gain at first use about $71,000 Derived, both computed below.

There’s a simple rule inside that arithmetic: a fixed gain in points is worth most at the stage with the lowest conversion rate, because it’s the largest relative lift there. Chapter 11 adds cost and confidence.

With the defaults, the stall is the second order: 1,950 customers a month stop there, with up to about $134,000 of contribution at stake per cohort. A 5-point gain at the second order is worth about $123,000 a year; first use loses nearly as many customers but the same gain there is worth about $71,000 Derived.

Read the leak carefully

• Look at three cohorts, not one. If the same stage leads in three in a row, it’s the stall. The Noise Floor shows how big a month-to-month change has to be before it means anything.
• Split by how the first order was acquired. A cohort bought on a 40%-off offer may leak at the second order for reasons no journey fix will cure; The First Offer covers that.

Do this

• Build the cohort table for the last three cohorts whose third-order window has closed, using the query in Appendix A. Put the counts into the tool. Write down the stall, and the runner-up if it’s within 10%.