Customer Metrics
How customers join, stay, pay and leave
Customer metrics — acquisition, activation, retention, churn, LTV and cohort economics — describe the health of the customer base. Cohort analysis ties them to time so trends are visible.
What you will learn
- Define acquisition, activation, churn, retention and LTV precisely
- Run a cohort analysis to make retention trends visible
- Relate LTV and CAC to decide if growth is sustainable
What is this?
Customer metrics measure the health of the customer base: acquisition (how many join), activation (who reach value), retention/churn (who stay or leave), revenue per customer, and lifetime value (LTV). Cohort analysis groups customers by signup period so you compare like-with-like over time instead of averaging across vintages.
Why do you need it?
- Company-wide averages hide the real story: today's retention can be worse while the average lag blurs it.
- Churn is a leading indicator of revenue loss; LTV vs CAC determines whether growth is profitable.
- Cohorts make seasonality and product changes visible, not mysterious.
- Customer metrics are the input every forecast and growth model depends on.
Mental model
A restaurant's table turnover and repeat diners
Customer metrics
New covers (acquisition), the moment a diner orders a second time (activation), returning guests (retention), and lifetime spend (LTV). A packed Tuesday that never brings people back is churn hiding behind a full room.
How it works
- Acquisition counts new customers in a period from a well-defined source (not every touch).
- Activation measures the fraction reaching a value moment (first purchase, first workflow).
- Retention tracks the share of a cohort still active after k periods.
- Churn is the complement on the right denominator: churned / due-for-renewal (not churned / total).
- LTV sums expected value per customer using retention curves and margin; CAC is acquisition cost per customer.
The customer lifecycle and its metrics.
In production
- A subscription app prints a cohort table (signup month × months since signup) so the team sees retention improving or decaying per vintage.
- An e-commerce team computes LTV by cohort and CAC by channel to cut spend on channels that buy unprofitable customers.
- A CS org defines at-risk customers by declining engagement cohort and routes them to save-play workflows.
Best practices
- Define activation as one observable value moment and measure its velocity.
- Choose churn's denominator deliberately: due-for-renewal, not total base.
- Use cohorts by signup (or first-purchase) period and report age-matched retention.
- Pair LTV with CAC and a payback window — LTV alone means little.
Common mistakes
- Averaging retention across cohorts and declaring 'retention is fine' while new cohorts decay.
- Computing churn as churned/total, which understates churn in a growing base.
- Counting re-signups as new customers, double-counting the base.
- Ignoring expansion in LTV and under-valuing retained growth.
Common issues
Retention looks flat but growth deceleratesCohort averaging masks newer-cohort decay
Report age-matched cohort retention and compare vintages at equal age.
Churn is 2% but revenue keeps fallingDenominator includes all customers, hiding heavy churn of big accounts
Churn by revenue and by count, with due-for-renewal denominators.
Debugging tips
- If cohort tables look noisy, check the cohort assignment rule (signup vs first order vs activated).
- If churn suddenly drops, check the denominator change, not just the numerator.
- Trace a customer from cohort row to event row to verify the counting logic end-to-end.
Interview checkpoint
Hands-on lab
- Assign customers to signup cohorts
- Compute age-month retention per cohort
- Diagnose a retention trend that averages would hide
Build a Cohort Retention Table in SQL
40 min · intermediate
Quick quiz
Which cohort metric most directly surfaces today's product problems?
Churn ÷ total customers is the correct churn formula.