SI Signal & Funnel
Growth Experiments

Cohort Analysis for Beginners: Keep Time Aligned

Signal & Funnel · Updated 2026-07-19 · 4 min read
Cohort Analysis for Beginners: Keep Time Aligned
SummaryBuild a cohort analysis by grouping a defined entity around a shared starting event, then comparing a separately defined outcome at aligned ages. Document time zone, exclusions, identity, reactivation, denominator, partial periods, and future unavailable cells. Show counts with rates, validate joins and delayed data, protect small or sensitive groups, and mark product or campaign changes. Treat cohort differences as patterns requiring explanation, not proof that one intervention caused them.
Contents

Group entities by a shared starting event

Cohort analysis follows a defined group that shares a starting event—such as first purchase, activation, signup, or acquisition period—then compares later behavior at aligned ages. It prevents a mature group and a brand-new group from being judged on the same calendar without acknowledging that one has had more time to do things.

Choose the cohort event from the decision, not from whichever timestamp is easiest to query.

Define the entity and start

State whether the entity is a person, account, payer, household, subscription, or organization. Define first event, time zone, exclusions, duplicates, migrations, reactivations, and what happens when identities merge or split.

Use the retention guide to define the later active or return event separately. “Signed up” and “used the core service” are different cohort starts.

Choose calendar and age periods

Group cohorts by a meaningful calendar interval, then measure behavior at age zero, one, two, and onward using a declared period such as day, week, or month. The appropriate period depends on natural usage and decision cadence; there is no universal choice.

Decide how partial periods are handled. Do not compare a completed third month with a third month that has only just begun.

Build the cohort table

Place acquisition cohorts in rows and aligned ages in columns. Show the starting count and the count or rate meeting the defined outcome at each age. Keep unavailable future cells distinct from zero.

For a rate, divide qualifying entities at an age by the documented eligible cohort base. State whether the denominator remains the original cohort or changes for legitimate eligibility rules.

Check data before reading color

Validate the start event, later event, identity, joins, time zones, delayed imports, refunds, deleted accounts, and backfills. Heatmap color can visually overstate tiny cohorts, so show counts and uncertainty.

The funnel guide helps when the question concerns sequential stages before the retained behavior.

Segment only with a reason

Compare acquisition source, plan, market, product, or another lawful segment when it tests a declared explanation. Preserve acquisition definition and aligned age. Do not test repeated segment cuts until one happens to look favorable; declare the comparison first.

Small or sensitive groups require privacy and fairness controls. Do not infer protected traits or expose individual behavior. Follow current privacy, consent, security, contract, and sector requirements under the relevant official regulator guidance named below and qualified local counsel.

Separate composition from treatment

A later cohort can perform differently because its audience, offer, price, onboarding, season, product, or tracking changed. Cohort analysis describes the pattern; it does not isolate which factor caused it.

Use experiments or appropriate causal methods for treatment effects. Mark releases and campaigns directly on the timeline.

Connect cohorts to economics carefully

The CAC guide can align acquisition cost with the customers who entered together. Avoid comparing complete cost with an immature cohort's partial return or retention.

Report the cohort difference, maturity, size, data breaks, explanations, and next test. Do not infer a cause from the cohort table alone.

Preserve cohort maturity

Freeze the acquisition rule and show the maximum fully observed age for each cohort. Gray out or label incomplete cells instead of comparing them with mature periods. When late events arrive, note the refresh date and whether the decision changed after maturation.

Official rule sources

Data-protection and direct-marketing duties depend on jurisdiction, data, purpose, and message. Check the current official source relevant to the people and activity: the European Commission data-protection portal for EU scope, the UK Information Commissioner's Office direct-marketing guidance updated 28 April 2026, the California Privacy Protection Agency laws and regulations for California scope, and the U.S. Federal Trade Commission CAN-SPAM guide for U.S. commercial email. These official pages do not determine whether a rule applies to a specific business. Also check current platform documentation and contracts, and use qualified local privacy or legal counsel for consequential decisions.

General marketing education, not legal, privacy, tax, financial, security, or individualized business advice. An independent publication. Not affiliated with any prior owner of this domain.

FAQ

What is a cohort in marketing analytics?

A cohort is a defined group of entities sharing a starting event or period, such as accounts first activated in the same month. The entity, start event, time zone, exclusions, and identity rules must be explicit. Cohorts allow later behavior to be compared at the same age rather than mixing mature and newly acquired groups in one calendar total.

How do I calculate cohort retention?

For each aligned age, divide entities meeting the defined return or active condition by the documented eligible cohort base, then multiply by 100 for a percentage. State whether the base is fixed or changes under legitimate eligibility rules. Keep future unavailable periods separate from zero, show counts, and align identity, time zone, and delayed validation.

Does a better cohort prove a campaign worked?

No. Cohorts may differ in audience, price, product, season, onboarding, acquisition source, tracking, or maturity. The table describes aligned behavior but does not isolate one cause. Mark known changes, investigate composition, and use an appropriate experiment or causal method for incremental claims. Report small cohorts and uncertainty rather than relying only on heatmap color.

Which official privacy and marketing sources should I check?

Use the official source that matches the people, jurisdiction, data, and activity: the European Commission data-protection portal for EU scope, the UK Information Commissioner's Office direct-marketing guidance for UK scope, California Privacy Protection Agency laws and regulations for California scope, and the U.S. Federal Trade Commission CAN-SPAM guide for U.S. commercial email. Then check current platform documentation and contracts. Qualified local counsel should review consequential decisions.