Products Solutions Templates Pricing Resources

Advanced segmentation and cohort analysis: understanding customer groups over time

Home / Everything About / Everything About Analytics / Advanced segmentation and cohort analysis: understanding customer groups over time

Segmentation answers: do different customers behave differently.

Example: customers who adopted feature A have two percent churn. Customers who did not adopt feature A have four percent churn. Feature A drives retention. Feature A is important.

Cohort analysis answers: do different groups of customers behave differently over time.

Example: customers who signed in January have eighty percent retention in month one, sixty percent in month two, forty percent in month three. Customers who signed in February have seventy-five percent retention in month one, fifty percent in month two. January cohort retains better. Something changed between January and February.

When to use segmentation

Use segmentation when metric varies by segment

Revenue varies by customer size (enterprise vs self-serve). Traffic varies by traffic source (paid vs organic). Engagement varies by product usage (power users vs occasional users). Segment and measure separately. Understanding variance improves decisions.

Do not segment if metric is same across segments

Page views are same on Monday and Tuesday. No point segmenting by day. Time is not a useful segment.

When to use cohort analysis

Use cohort analysis when cohort matters

How does customer lifetime value differ by sign-up month. Months matter (seasonality, product changes). Cohort analysis reveals time effects.

Do not use cohort analysis if cohort is meaningless

Pages loaded per visit does not vary meaningfully by sign-up month. Cohort analysis would show noise, not signal.

Real example: advanced segmentation at SaaS company

Initial segmentation by size

Enterprise (annual contract value one hundred thousand plus): twenty customers. Two percent monthly churn. Mid-market (ten thousand to one hundred thousand): one hundred customers. Four percent monthly churn. Self-serve (under ten thousand): five hundred customers. Eight percent monthly churn.

Enterprise customers are sticky. Self-serve customers churn fast. This is expected. But mid-market churn at four percent is above target of three percent.

Further segmentation by behavior

Further segment mid-market by product adoption. Mid-market customers using feature A (adoption): two percent monthly churn. Mid-market customers not using feature A (no adoption): six percent monthly churn.

Feature A is retention driver for mid-market. Action: improve feature A discovery for mid-market customers. Get more mid-market customers to adopt feature A. Churn should decrease.

Real example: cohort analysis at marketplace

Track buyer repeat rate by sign-up cohort

January cohort (one thousand buyers): repeat rate month one is thirty percent. Month two is twenty percent. Month three is fifteen percent.

February cohort (one thousand buyers): repeat rate month one is twenty-five percent. Month two is eighteen percent. Month three is twelve percent.

Discover difference

January cohort retains better than February. Why. Investigate.

Test hypothesis

Hypothesis: January had special onboarding. February did not. Test: reintroduce onboarding for March cohort.

March cohort (one thousand buyers): repeat rate month one is thirty percent (similar to January). Month two is twenty-two percent. Month three is sixteen percent.

Onboarding improved retention. Feature A was the fix.

Segmentation and cohort analysis together

Combine both for powerful analysis

Segment enterprise customers by cohort.

January enterprise cohort: eighty-five percent renewal rate. February enterprise cohort: eighty percent renewal rate. March enterprise cohort: ninety percent renewal rate.

Investigate and act

March enterprise is renewing better. Why. Investigate. Find: March cohort was onboarded with new process that emphasizes relationship building. Relationship building increases enterprise renewal rates. Action: apply new onboarding to all cohorts going forward.