What is bounce rate, and how do you measure it in Metabase?

Bounce rate is the share of website sessions with no engagement — in GA4 terms, sessions that weren't engaged: under 10 seconds, no conversion event, one pageview. It's the fastest read on whether a landing page and its traffic source actually match. (Looking for the email deliverability metric? That's email bounce rate — a different number entirely.) Measure it in Metabase from session data exported from Google Analytics, Plausible, or PostHog.

What does a bounce rate chart look like in Metabase?

Track one landing page's bounced-session share by month and read it downward: the slide from about 62% to 48% is what page and message fixes look like compounding over a year. A one-month spike like April's usually means a new traffic burst that didn't match the page's promise — segment that month by source and medium before touching the page itself.

What bounce rate measures

It measures mismatch. A bounce means the promise that brought someone to a page — an ad, a search snippet, a social post — wasn't kept fast enough for them to do anything at all. Segmented by landing page it finds pages that underdeliver; segmented by source it finds traffic that was never going to engage. As a site-wide average it measures nothing, because it blends blog readers who found their answer with checkout visitors who fled, and feeds neither engagement rate analysis nor conversion work.

GA4 vs. the old definition

Universal Analytics called any single-pageview session a bounce, however long it lasted. GA4 inverted the frame: it defines engaged sessions — longer than 10 seconds, or containing a conversion event, or with two-plus pageviews — and a bounce is simply a session that wasn't engaged. The practical consequences: single-page content that holds attention no longer bounces, reported bounce rates fell sharply on migration without any behavior change, and any trend that crosses the migration date needs a seam drawn on the chart. If you compute bounce rate yourself from raw events, write the engagement rule into one SQL model so every card shares it.

What data does it need?

• A sessions table (or a model built from raw events) with session_start, landing_page, source, medium, and an is_engaged flag computed from duration, conversions, and pageview count.
• A converted flag per session if you want bounce and conversion side by side — the pairing that makes the metric actionable.
• Source: the GA4 BigQuery export, or self-hosted analytics events from Plausible, PostHog, or Mixpanel landed in the warehouse by a pipeline tool.

SQL patterns

SELECT
  landing_page,
  date_trunc('month', session_start) AS month,
  COUNT(*) AS sessions,
  ROUND(
    100.0 * COUNT(*) FILTER (WHERE NOT is_engaged)
    / NULLIF(COUNT(*), 0), 1
  ) AS bounce_rate_pct
FROM sessions
WHERE session_start >= CURRENT_DATE - INTERVAL '6 months'
GROUP BY 1, 2
HAVING COUNT(*) >= 100
ORDER BY 1, 2;

SELECT
  source,
  medium,
  COUNT(*) AS sessions,
  ROUND(
    100.0 * COUNT(*) FILTER (WHERE NOT is_engaged)
    / NULLIF(COUNT(*), 0), 1
  ) AS bounce_rate_pct,
  ROUND(
    100.0 * COUNT(*) FILTER (WHERE converted)
    / NULLIF(COUNT(*), 0), 2
  ) AS conversion_rate_pct
FROM sessions
WHERE session_start >= CURRENT_DATE - INTERVAL '30 days'
GROUP BY 1, 2
HAVING COUNT(*) >= 200
ORDER BY sessions DESC;

Pitfalls

Where this metric applies

• Google Analytics + Metabase — engaged-session data via the BigQuery export
• Plausible + Metabase — lightweight session stats by page and source
• PostHog + Metabase — raw events to build your own engagement rule
• Mixpanel + Metabase — session and event exports for engagement models

Related

Metrics

Dashboards

FAQ