Unit economics
Cohort analysis
Grouping customers by when they arrived and following each group separately, so that changes over time are visible instead of averaged away.
A cohort is everyone who started in the same period. Following them separately is the only way to tell whether the product improved or the mix changed: a company that starts buying cheaper traffic will show worsening blended retention even if every cohort behaves exactly as before, and a company that improved its onboarding will show it in the newest cohorts first while the blended number lags for months.
The table is read in two directions. Down a column is the effect of tenure — how any cohort behaves at month three. Across a row is the effect of calendar time — what happened to everyone in March. A number that moves down a column is a product fact; a number that moves across a row is usually an operational one.
Blended metrics change whenever the mix changes, which on a growing business is constantly. Most reported improvements in retention are composition effects, and most reported declines are the arrival of a cheaper channel rather than a worse product.
Knowing the definition is not the same as being able to check the figure. These are the procedures that do the second thing.
- Testing a growth number you were given
- “Conversion is up 6.4×. Revenue grew 40%. Abandoned carts down 65%.” · 30 minutes, 5 questions.
A definition is free. Being answerable for the figure it produces is the part that is bought, and this term is a working part of the engagements below.
- Strategy and market research
- State the assumption that would change the answer, then go and test that.
- Client-base reactivation
- The cheapest customer already gave you their contact details.
- Business process audit
- Remove before you optimise. Most process debt is subtraction, not addition.
The definitions are the easy part. Whether the figure on your dashboard was computed this way is a different question, and usually the more expensive one.