Fictional example: 35% bounce rate
Within one measurement scope, 200 sessions include 130 engaged and 70 non-engaged sessions. Engagement rate is 130 / 200 = 65%; bounce rate is 70 / 200 = 35%. The two rates sum to 100% and describe the same session population.
| Fictional measure | Value | Calculation or definition |
|---|---|---|
| Sessions | 200 | Same period and scope. |
| Engaged sessions | 130 | Applicable GA4 criteria. |
| Non-engaged sessions | 70 | 200 − 130. |
| Engagement rate | 65 % | 130 / 200 × 100. |
| Bounce rate | 35 % | 70 / 200 × 100. |
Check criteria before comparing
- A session is engaged if it exceeds the applicable time threshold, has a key event or at least two page or screen views. The web time threshold defaults to 10 seconds and can be adjusted.
- Record the threshold, key events, filters, period and collection scope. Configuration changes can move the rate without improving a journey.
- Compare similar pages and intentions. Finding an answer quickly can be successful; a poorly defined key event can instead flatter the score.
Check the person’s intended outcome
Find a reproducible difficulty and compare the rate with task completion, errors and user feedback. Do not create unnecessary events to reduce bounce rate. An observed change establishes neither its cause nor a business gain.
Primary references
- Google Analytics — Engagement rate and bounce rate [EN]
- Google Analytics — Sessions and settings [EN]
GA4 bounce depends on the engaged-session definition
In GA4, bounce rate is the proportion of non-engaged sessions and complements engagement rate. A scroll event does not automatically qualify a session. Document configuration, period and unit before comparison, especially after changing key events or collection.
- Check applicable duration, key events and page or screen views.
- Compare the same reports and populations.
- Separate measured engagement from user-task success.
Does high bounce prove a poor page?
No. A brief answer can satisfy a need quickly. Examine task, content and collection rather than reducing bounce by adding artificial events.
Primary reference: Google Analytics — Engagement and bounce rate. Related method: read the practical guide.
Engagement: add counts before rates
A lower bounce rate is interpretable only with comparable engaged-session definitions and collection. The reconciliation below shows why averaging percentages can produce a misleading interpretation.
Fictional teaching case — figures describe neither CIE nor a client.
In this complete fictional extract, engagement rules, period and property are identical across devices. Desktop has 800 sessions with 480 engaged; mobile has 200 sessions with 60 engaged. Their simple average, 45%, is not the combined rate of 54%.
- Combined rate
- 54%540 / 1,000 sessions
- Simple average of rates
- 45%(60% + 30%) / 2
- Combined bounce rate
- 46%1 − 54%, with matching definitions
| Fictional device | Sessions / engaged | Engagement rate |
|---|---|---|
| Desktop | 800 / 480 | 60% |
| Mobile | 200 / 60 | 30% |
| Combined | 1,000 / 540 | 54% |
- Desktop60%
- Mobile30%
- Weighted combined54%
Checks for interpreting the case
- Record property, time zone, period, filters and engagement rule. Collection changes can cause a difference without a changed journey.
- Keep sessions, users and events distinct. An event firing several times does not become several engaged sessions.
- For overlapping dimensions, check that a session does not appear in multiple rows before adding them. A missing value is not zero.
Check the denominator before explaining the decline
A changing audience mix can shift an overall rate without a decline inside each segment. Keep the counts and collection changes alongside the comparison.
| Signal to examine | Concrete check | Effect on the decision |
|---|---|---|
| Growing mobile share | Recalculate device rates using the same session definition. | Separate composition effects from a local decline. |
| Changed tag or consent | Look for a collection break at the change date. | Mark the series as non-comparable while scope differs. |
| Engaged sessions, abandoned task | Compare engagement with the business confirmation within authorised measurement. | Review the journey; engagement is not proof of task completion. |
From a number to a verification
Separate real change from a measurement change
- Test to reproduce
- Fix the period, time zone, device and engaged-session definition. Check whether a tagging change alters the denominator between periods.
- Stop signal
- Treating an increase after a collection change as evidence of a better journey.
Change an assumption
A better total can hide declines on every device
Segment volumes can move the overall rate without improving the journey. These figures are a second fictional example, separate from the preceding table.
- Period A
Desktop: 60/100 engaged sessions; mobile: 90/300. Total: 150/400, or 37.5%.
- Period B
Desktop: 165/300; mobile: 25/100. Total: 190/400, or 47.5%.
- Interpret
The total gains 10 percentage points, while desktop falls from 60% to 55% and mobile from 30% to 25%.
Reading exercise: the variations describe fictional situations.
Can we claim a better experience?
- New evidence
- The share of desktop sessions rises from 25% to 75%.
- What changes
- The composition change explains the higher total in this example. Examine device-level rates before concluding; both have declined.
What if the engagement definition changed?
- New evidence
- An ordinary event becomes a key event between periods.
- What changes
- Document this measurement break. A comparison under a constant definition is needed before interpreting the change as different behaviour.
Is bounce rate always the complement of engagement?
Under the cited GA4 definition, yes: it concerns non-engaged sessions. The relationship requires identical periods, filters and session sets.
Related method: read the practical guide.