94% Bounce Rate Was a Complete Lie — Here's What Was Really Happening

The situation

A UK fashion brand's GA4 account showed a 94% bounce rate and an 8-second average session duration site-wide. Their marketing director had presented this data in a board meeting and leadership had pulled budget from several channels as a result. All data-driven decisions were on hold.

The problem

Three completely separate GA4 configuration tags were active inside their GTM container simultaneously — a tag from the original setup in 2022, one added by a freelancer in 2023, and a third added by their internal developer without realizing the others existed. Every single pageview was being counted two or three times. When GA4 sees the same session start event fire multiple times in quick succession, it creates multiple sessions — which appear to start and immediately end. This artificially inflated total sessions, made every session appear to last seconds, and made bounce rate look catastrophic.

What we did

1. Full GTM container audit — catalogued all active tags and traced each back to who added it and when
2. Identified all 3 GA4 configuration tags and confirmed they were all linked to the same GA4 property
3. Removed the 2 duplicate configuration tags, leaving a single clean implementation
4. Rebuilt the full event schema from scratch — pageview, scroll, add_to_cart, begin_checkout, purchase — and verified every event in GA4 DebugView before launch
5. Ran a 7-day post-fix comparison to confirm data was now consistent with what the team expected from site traffic

Why it worked

GA4 session logic is sensitive to duplicate configuration tags in a way that isn't immediately obvious. When two config tags fire, GA4 creates two session_start events — which GA4 interprets as two separate sessions of near-zero duration. This is a common consequence of multiple agencies touching the same GTM container without a proper handover audit.