A single site-wide bounce rate hides where you're actually losing people. Cohort Bounce breaks bounce down two ways at once — by landing page and by acquisition cohort (channel or device) — so "42% bounce" becomes "paid traffic to /pricing bounces at 85%, but organic bounces at 9%".
Where to find it
Open a website, go to the Retention tab, and switch the top toggle from Retention / LTV to Bounce. It's part of the Retention tab, so it needs a Pro plan or higher. It does not appear on public shared dashboards.
Anatomy of the grid
- Rows — your top 20 landing pages by sessions for the selected date range (the first page of each session).
- Columns — the cohort: traffic channel (Direct, Organic, Paid, Social, Referral, …) or device (Desktop, Mobile, Tablet, …). Use the Channel / Device toggle in the card header. Up to 9 columns, ordered by volume.
- Cells — the bounce rate for that page × cohort pair. Colour scales with bounce (pale → deep red; legend at the bottom). An
n/achip means there were no sessions for that pair.
Hover any cell to see its session count, bounce %, and (in the card footer) how far that rate sits from the page's own average.
What a "bounce" is here
A bounce is a single-page session — one pageview, no second navigation. This is the same definition the Pages table uses, so the numbers reconcile: a page's overall bounce equals its Cohort Bounce row combined across cohorts.
Biggest bounce leaks
The right-hand Biggest bounce leaks rail ranks the worst page × cohort pairs so you don't have to scan the grid. It only lists pairs with at least 5 sessions — a single-session "100%" is noise, not a leak, and surfacing it would erode trust in the list. Click any leak to jump straight to those sessions.
Acting on it
Find the deep-red outliers
Scan for cells far darker than the rest of their row. A page that's pale for organic but deep red for paid means the paid landing experience (ad copy ↔ page mismatch, slow load, wrong offer) — not the page itself — is the problem.
Use the leaks rail to prioritise
The rail is already ranked by severity and filtered to ≥5 sessions. Start at the top.
Drill into the sessions
Click a cell or a leak. Zenovay applies the landing-page and channel/device filters and switches to the analytics view, pre-filtered to exactly those sessions, so you can see what those visitors did.
Why channel drill-down is labelled "approximate"
Device is exact, channel is approximate
The grid groups by the channel stored on each visitor at collection time. The session filter you land on after a click re-derives channel from UTM tags, referrer, and click-IDs. These two methods agree the large majority of the time but can differ by a few percent, so the grid honestly labels channel drill-down as approximate. Device grouping and drill-down are exact.
Limits
| Limit | Value |
|---|---|
| Landing pages (rows) | Top 20 by sessions |
| Cohort columns | Up to 9, by volume |
| Leaks rail minimum | 5 sessions per pair |
| Plan | Pro and above (part of Retention) |
Lower-traffic pages and rarer cohorts fall below the cut so the grid stays legible and fast. Pick a longer date range to pull more pages into the top 20.