Funnels, cohorts and segmentation
Where is the important loss, and for whom?
How I use it
Map a useful journey
Choose the start and the value event, then define the steps, order and time allowed between them. A high-level funnel may be discover → try → reach value → return → pay. A nested funnel can unpack one difficult step. Real journeys may branch, so check that the map fits the product.
Use consistent denominators
Distinguish unique people from events, eligible customers from all customers, and account adoption from seat activity. Inspect missing events and identity joins.
A falling conversion rate can reflect a change in who enters the funnel rather than a worse experience.
Compare meaningful segments and mature cohorts
Look at source, role, account size, use case, geography and tenure where you have a reason to expect a difference. Follow groups from a common starting event and give them equal time to mature.
Use both rates and population size to judge whether a difference matters.
Turn a pattern into a question
Combine behaviour with interviews, support themes, sales win-loss evidence or observed usage.
A segment difference is a clue, not proof of cause. Write the competing explanations and the cheapest useful way to distinguish them.
An illustrative funnel
- Eligible trial accountsStarting population1,000
- Start setup60% of eligible accounts600
- Complete setup50% of setup starts300
- Finish useful work40% of setup completions120
Illustrative trial accounts: 1,000 eligible → 600 start setup → 300 complete setup → 120 finish useful work. Counts and conversion tell different parts of the story.
Of 1,000 eligible trial accounts, 600 start setup, 300 complete it and 120 complete a useful workflow. The largest loss by count is before setup; the weakest step conversion is from setup completion to useful work (40%). Neither automatically deserves first priority.
If research shows that many visitors are a poor fit, fixing the last step for high-intent teams may be worth more. Opportunity size depends on the reachable population, likely effect and downstream value, as well as the visible drop-off.
What to watch for
- Picking the biggest percentage drop without checking value and population.
- Comparing a new cohort at day seven with an older cohort at day thirty.
- Searching dozens of tiny segments until an exciting difference appears.
A brief for the next discussion.
A specific customer group and journey step to investigate, with a plausible explanation to test.
Find the constraint
Where is your growth work getting stuck?
Use the growth diagnostic to find where to focus, or talk it through with me.