Cohort metrics: compare retention, revenue and contribution

Cohort metrics compare retention, orders, revenue and contribution for a fixed group of customers at the same age. The useful question is whether a newer acquisition group returns and earns more after comparable time has passed. Revenue alone cannot answer it, because a few large orders can conceal fewer returning accounts and lower contribution.

Define the membership and the clock

A cohort needs an entry rule, an activity rule and a reporting clock. For a distributor, entry might be the first posted customer invoice. Activity might mean at least one completed, noncanceled order during a later calendar month. Choose the customer account as the unit, rather than mixing accounts with individual buyers.

Freeze the original membership. A January customer remains in January even after an inactive month. Do not admit a returning legacy account merely because its earliest invoice in a partial export happens to be in January; use the customer master or sufficient earlier history to establish first purchase.

Keep period activity distinct from cumulative activity. Google's cohort exploration documentation distinguishes activity within one period from activity accumulated across periods. The same distinction matters in a business export: customers ordering in the next month are a different measure from customers ordering at any time afterward.

In the example below, age zero is the acquisition calendar month and age one is the next complete calendar month. The data cutoff is March 31, 2026, so both January and February cohorts have completed age one. Age zero gives different exposure to early and late joiners; use fixed elapsed-day windows if that difference would change your decision.

The measures and their denominators

Use these definitions before choosing a chart:

Net revenue should reflect the credits, returns and discounts applicable to those sales. The SEC's guide to financial statements explains the distinction between revenue after returns and allowances, gross profit after cost of sales, and operating expenses. State which costs your contribution measure subtracts; it is not interchangeable with every other profit subtotal.

Here, variable costs include the goods sold and order-specific fulfillment costs. They exclude acquisition spending and fixed overhead. The separate gross margin and contribution margin guide explains the cost boundaries in more depth.

Work every measure from the same customer rows

An illustrative US industrial distributor supplies a customer roster, completed-order summary and finance cost summary. The following age-one rows include every original customer, including those who did not return. Money is USD; displayed percentages are rounded to one decimal place, using unrounded values for calculations.

Contribution on each row is revenue minus variable costs. January has three returning customers from four original accounts; February has two from four. Those same rows produce the scorecard:

February produces more revenue per acquired customer, but fewer customers return and less contribution remains. The sales owner should investigate customer concentration; finance should inspect discounts, product mix and fulfillment cost. The table identifies questions, without proving that an acquisition campaign caused the difference.

Prove the report before acting

First, tie each cohort's original count to its frozen roster, including inactive accounts. Then reconcile orders to the completed-order register and revenue and costs to the finance rows for those accounts and age windows. February's March activity must not be compared with January's entire history.

A useful identity checks the denominator chain:

Revenue per original customer = retention × orders per returning customer × revenue per order.

January is 0.75 × 2 × (24,000 / 6) = $6,000. February is 0.50 × 2.5 × (26,000 / 5) = $6,500. This ties the customer, order and revenue measures together; it does not replace the ledger reconciliation.

For a longer time series, use the existing cohort retention table guide. For future value estimates, follow customer lifetime value from the ledger rather than extending this observed month indefinitely.

Ownership and where it goes wrong

Revenue operations owns the roster and first-purchase classification. Sales or customer success owns the response to weak repeat activity. Finance owns the revenue and cost reconciliation. Record definitions, exclusions and effective dates using versioned metric definitions, so a changed rule does not look like changed customer behavior.

• Survivors replace the original denominator. Removing inactive accounts overstates retention and revenue per acquired customer.
• Younger groups get zeros. An unfinished age window is unavailable, not a completed period with no activity.
• Costs disappear on inactive rows. A customer with no order can still generate service costs; include them if the contribution definition requires them.
• Large accounts conceal breadth. Higher cohort revenue can coincide with fewer returning customers. Read counts and dollars together.

Covirage is not affiliated with Google.

Compare cohort metrics on your own file in Covirage

Bring a sample of your own customer roster, dated order export and finance cost summary to Covirage, which maps the columns. Its tools compute and check comparable cohort counts, revenue and contribution; the external AI model explains the results with every figure traced to its rows. Create a free Covirage account and bring a sample of your own data.

See Covirage for sales insights for the business context.