Bounce rate is the most watched and worst interpreted metric in web analytics. A high number triggers pointless redesigns; a "decent" number lulls pages that lose customers. Let's put it back in its place: a symptom to interpret, never a verdict.
Quick Answer: bounce rate in brief
Definition: the share of sessions where the visitor leaves without interaction — no click, no second page, no conversion.
- Benchmarks — 25–40% in e-commerce, 40–60% in B2B, 60–90% for blogs (and that's healthy: a served reader leaves happy).
- What it doesn't say — whether the bounce is good (answer found) or bad (page disappointed). To decide, cross it with time on page and scroll depth.
- SEO impact — not a direct ranking signal, but a bad bounce usually accompanies what Google actually penalises: the mismatch between page and search intent.
- Reducing it — align the page with its traffic source, load faster, make the next action obvious.
What "bounce" means depending on the tool
A classic trap: every tool has its own definition. Single-page-view session (old Universal Analytics), session without an "engagement event" longer than 10 seconds (GA4, inverted into "engagement rate"), session without meaningful click or scroll (behavioural tools). Consequence: the same site shows 55% here and 38% there, and neither number is wrong. Compare your pages against each other inside one tool — never your numbers against another ecosystem's benchmarks.
Good bounce, bad bounce: the reading grid
A bounce is not a failure per se. Four cases:
| Situation | Time on page | Scroll | Diagnosis |
|---|---|---|---|
| Flash bounce | < 10 s | None | Bad: broken promise, slowness, misleading ad |
| Reader bounce | > 1 min | Deep | Good: content consumed — normal for blogs, FAQ, contact |
| Hesitant bounce | 30–60 s | Partial + back-and-forth | Recoverable: real interest, unresolved friction or doubt |
| Utility bounce | < 30 s | Targeted (hours, address, price) | Good: express answer found |
The "hesitant bounce" is your goldmine: these visitors wanted something. Five session replays are usually enough to see what stopped them.
Diagnosing: per page and per source, never globally
A site's global bounce rate averages incomparable things — it rises when your blog performs, it falls when a mediocre campaign stops. The useful reading:
- Per landing page: rank your entry pages by traffic × bounce. A major landing page bouncing abnormally high for its category = priority.
- Per source: the same page can do 30% from Google and 85% from a social ad — the problem is then the ad's targeting or promise, not the page. Same principle as for a landing page.
- Per device: a mobile/desktop gap above 20 points almost always signals a mobile speed or ergonomics problem.
Reducing a bad bounce: the six levers
- Keep the source's promise. The page title must echo the intent that brought the click — Google query, ad, post. Message-page continuity is lever #1.
- Load fast. Every second beyond 2–3 s is paid in flash bounces, especially on mobile.
- Answer first, elaborate after. The answer to the intent at the top of the page (exactly the principle of our Quick Answers), the details below.
- Make the next step obvious. A served reader leaves… unless a relevant internal link or contextual CTA offers the logical continuation. Interlink.
- Clean up mobile readability. Line heights, font sizes, no aggressive interstitials.
- Verify with the heatmap: if the scroll map shows a massive drop before your first argument, move it up.
Bounce and SEO: what Google actually sees
Google doesn't read your analytics. But a bad bounce has a twin Google does observe: pogo-sticking — the quick return to the results page followed by a click on a competitor. Reducing flash bounces on your SEO landing pages mechanically reduces that negative signal. Conversely, sacrificing clarity to artificially "retain" (padded content, answer buried at the bottom) degrades the experience AND eventually costs you. Serve the intent, fast — the SEO module crossed with behaviour shows which queries bring visitors who stay.
In practice with Mirage — Engagement reads without configuration: duration, read rate and engagement rate per page, bounce by source and device, plus one-click access to bounced sessions in replay to see what happened — on 100% of traffic, banner or not. Free 30-day trial.
FAQ
What is bounce rate?
The share of sessions that end without meaningful interaction: the visitor lands on a page and leaves without clicking, navigating or converting. Beware of definitions: depending on the tool, a bounce may require zero clicks, less than N seconds, or a single page view — only compare numbers built on the same definition.
What is a good bounce rate?
By page type: 25–40% for e-commerce, 40–60% for a B2B site, 60–90% for a blog post — and that is normal: a reader who finds their answer and leaves is a satisfied bounce. A site-wide rate mixes everything; analyse per page and per source.
Does a high bounce rate hurt Google rankings?
Not directly — Google has repeatedly said your analytics tool's bounce rate is not a ranking signal. However, what bounce reveals (a page disappointing search intent) affects signals Google does measure: quick returns to results, no follow-up click. Fixing a bad bounce improves both.
How do I tell a good bounce from a bad one?
Cross it with time on page and scroll. Bounce + 3 seconds + zero scroll = the page disappointed. Bounce + 2 minutes + full read = content consumed with satisfaction (common on blogs). Session replay removes the last doubts by showing what these visitors actually do.