Case: a geo-aware CDN swap cut prelander bounce 21% in tier-3 markets

Case: a geo-aware CDN swap cut prelander bounce 21% in tier-3 markets

The gap: identical campaign, two geos. Tier-1 ran at 3.8% CR, a tier-3 geo at 1.9% on the same offer and lander. The angle wasn't the problem.

The diagnosis:
1. Pulled tracker bounce data by geo — tier-3 prelander bounce was 64% vs tier-1's 38%
2. Tested prelander load from a server in the tier-3 region: 5.2s to interactive
3. The lander pulled images and a script from a CDN with no edge node near that region
4. Bounce spiked past the 3s mark, exactly where load was stalling

The fix:
— Moved the lander to a CDN with edge presence in the target region
— Compressed the hero image from 480KB to 90KB (WebP)
— Inlined the small script instead of fetching it
— Re-tested load: 1.6s to interactive in-region

Outcome: tier-3 prelander bounce fell from 64% to 43% — a 21-point drop. CR rose from 1.9% to 2.8%, closing most of the gap to tier-1, on identical spend and creative.

When one geo underperforms on the same creative, suspect load time before the angle. Test from inside the target region.

Save this SOP. Run this before every launch.