A conversion funnel describes a chosen sequence of recorded actions toward a defined outcome. Its counts depend on who qualifies for each step and the observation window.
Before interpreting a decline between steps, establish what the counts represent. A missing recorded action can tell you where to investigate, but it does not by itself reveal why someone stopped or whether they returned later.
Write down what qualifies for each step
Name the action, the people eligible for the analysis and the observation period. Then specify the entry step, the required order and whether other actions may occur between steps. Use the same counting unit throughout the sequence.
A recorded purchase is an analytics observation. Check what fires that event and whether it corresponds to the business outcome you need. A repeated event, a submitted form and a confirmed order answer different questions; do not silently substitute one for another.
A fictional product-to-purchase record
Imagine a shop reviewing one hypothetical reporting week. It counts distinct observed users who view a product, then start checkout, then have a purchase recorded. Each person is counted once at each reached step. Intervening actions are allowed; all included actions must occur within that week. These are the worksheet’s chosen rules, not independently collected shop data.
| Recorded step | Hypothetical users | Rate and denominator |
|---|---|---|
| Product view, entry | 120 | Entry population; no preceding step in this worksheet. |
| Checkout after product view | 30 | 30 ÷ 120 × 100 = 25% of entry users. |
| Purchase recorded after that checkout | 18 | 18 ÷ 30 × 100 = 60% of checkout users; 18 ÷ 120 × 100 = 15% of entry users. |
Checkout users are included in the entry population, and purchase users are included in the checkout population. Someone who first appears at a later step is outside this particular record. Activity after the reporting week is also outside it, even if the person later completes a purchase.
Read the percentage with its denominator
The 60% rate answers how many of the checkout users subsequently had a purchase recorded. The 15% rate uses everyone at entry instead. Neither is the other with a different label. Write the numerator and denominator beside the percentage, as the Conversion Rate definition explains.
These rates are not categories to add into a total. A person who reaches purchase is already included in the earlier counts. The week is a reporting boundary, so a missing later observation does not establish that someone permanently abandoned the purchase.
What GA4’s configuration changes
Google Analytics 4 (GA4) distinguishes open funnels, which allow entry at any step, from closed funnels, which require entry at the first. It counts users in the specified order and reports the first sequence under those rules within the date range. The setup also controls whether other actions may intervene between steps.
Record those settings when comparing results. If the entry rule, step conditions or observation period changes, describe the difference before treating the counts as the same measure. The familiar shape of a chart does not make its populations comparable.
Investigate the observation before naming a cause
For the fictional shop, check whether the intended actions produced the expected records and whether the same eligibility rules were used at every step. Review the date boundary too: someone entering near the end has less time to complete the sequence within that reporting week.
If the records appear sound, identify a specific unanswered question about the experience. You might review the checkout instructions or ask people attempting the task what was unclear. Those are ways to investigate a hypothesis; the count table alone has not identified the cause or established which change would help.
Keep the definitions with the decision
Save the action definitions, population, ordering rule and reporting window with the counts and calculations. Note any unresolved event-quality issue before choosing a comparison. The next review should make clear whether the measure stayed the same and what additional evidence supports a proposed change.
Google’s documentation supplies the GA4 configuration rules. The shop and its counts are fictional.