User funnel: define stages and compare cohort conversion

A user funnel measures how a defined cohort progresses through recorded website or product stages, using the same identifiers and observation window throughout. It tells you which transition needs investigation, provided the counts represent ordered actions by the same users rather than unrelated activity totals.

Define the unit before defining success

Start with a business question: do new purchasing users reach the action that makes an account useful? A website funnel might follow account creation, verification, a catalog search, and a saved purchasing list. Those are recorded user actions, not sales representatives moving opportunities through a CRM.

Choose one stable, de-identified user identifier. An account shared by several purchasing employees is a different unit from an individual user. If the business question concerns account adoption, build an account funnel rather than adding user counts and calling them companies.

Also distinguish users from events. One employee might search the catalog repeatedly. That activity belongs to one user at the search stage, regardless of how many searches occurred. Google's documentation distinguishes total, new, active, and returning users; those labels refer to different measures and should not be treated as interchangeable export columns (Google's user metrics guide).

Write a stage contract

For each stage, record the qualifying action, the identifier, and the deadline. State whether a user must complete every earlier stage in order. In a closed funnel, later activity does not qualify someone who never met the starting condition.

Google's funnel exploration documentation distinguishes open entry from closed entry and explains how specified event sequence affects inclusion. You need the equivalent definitions in your supplied export even when the analysis is performed elsewhere.

For the example below, a fictional US industrial supplier sets these rules:

Treat intervening actions as permitted. A person can browse help between verification and searching without leaving this funnel. A search recorded before verification does not qualify unless a later search meets the sequence rule.

The product funnel analysis guide applies this framework specifically to signup-to-activation diagnosis. Commercial handoffs between marketing, sales, and customer success belong in revenue operations metrics.

Measure mature cohorts from the supplied rows

Bring an event-summary CSV containing the de-identified user ID, signup timestamp, cohort week, qualifying stage timestamps, and export cutoff. The person preparing the file must preserve missing timestamps as missing. A blank cannot prove that an action never happened outside the recorded system.

Compare users who signed up in the weeks beginning September 7 and September 14, 2026. Extract the file on September 28, after everyone in both weekly cohorts has had the full seven-day observation window. Include only actions completed within each user's individual deadline, even if later actions are already visible in the file.

The resulting counts are:

Stage conversion = Users reaching the next stage / Users reaching the preceding stage

Overall conversion = Users completing the final stage / Users entering the first stage

Apply those divisions to the counts, retaining unrounded values for calculations and displaying percentages to one decimal place:

Combined conversion divides combined counts. Averaging the displayed cohort percentages would give each cohort equal weight despite their different sizes.

Prove that the counts belong together

Subtract adjacent counts to get users who had not reached the next qualifying stage by their deadline:

The check is 150 + 144 + 111 + 135 = 540. Each user occupies exactly one terminal bucket at the deadline. Also test the underlying records: every saved-list user must appear among qualifying searchers, and every searcher among verified users. Matching aggregate totals alone cannot prove those memberships.

If the export contains only summary counts, ask its owner to confirm the sequence and deduplication rules. An analyst cannot reconstruct ordered individual journeys from four independent event totals. Preserve the definitions alongside the file; analytics data portability explains why exported figures need their meaning attached.

Turn the result into an owned investigation

The later cohort's overall conversion fell from 30.0% to 21.0%, a decline of 9.0 percentage points. Every transition weakened, so there is no basis for blaming only the final action. Review verification delivery with the account operations owner, catalog access with the website owner, and list-saving feedback with the product owner.

The analyst owns count reconciliation. The product or website owner owns the stage definition and investigation. Customer support supplies evidence about the reasons people stopped. Compare acquisition source and account type before interpreting the cohort difference as an effect of a recent change.

Where it goes wrong

• Mixed units. Events, people, and company accounts produce different denominators. Keep one unit through the sequence.
• Unequal follow-up. A recent cohort with incomplete observation time will appear weaker even if behavior is unchanged.
• Independent stage totals. People searching before verification cannot automatically be counted as progressing through this ordered funnel.
• Unrecorded actions treated as confirmed exits. Missing events, blocked access, and identifier changes require review before a business explanation.
• A difference treated as proof. Cohort conversion is descriptive. A controlled evaluation is needed to attribute a change to a specific intervention.

Covirage is not affiliated with Google.

Compare your own user cohorts in Covirage

Upload a prepared event-summary file or Google Sheet with stage definitions and a consistent observation window. Covirage maps the columns, and its tools compute and check the cohort counts, conversion rates, and reconciliations; 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 sales insights for the business questions those file-based comparisons support.