Case: a 40-second postback window was inflating conversions by 18%

Case: a 40-second postback window was inflating conversions by 18%

Setup: Keitaro, push traffic, S2S postbacks from one network firing twice on confirmed leads (init + approve, both carrying status sale).

What we measured first:
— Pulled 7 days of raw postback logs, grouped by subid
— 11.4% of conversions had a duplicate within 60 seconds
— Reported CR was 4.7%, real CR was 4.0%

What we changed:
1. Enabled conversion deduplication on subid in Keitaro (Settings, Conversions)
2. Set the dedup window to 600 seconds, not 60 — the network's retry batch came at the 4-minute mark
3. Mapped the network's two events to distinct statuses: init then approve, so only approve counts as a sale
4. Re-imported 7 days to confirm the new count

Outcome:
— Conversions dropped 11%, CR corrected from 4.7% to 4.0%
— Two campaigns we were scaling on the inflated number turned out to be break-even, not 1.4 ROI
— Killed both before the next $600 spend

The lesson: a dedup window shorter than the network's retry interval is worse than no dedup. Check the retry interval in the logs, then set the window above it.

Save this SOP. Run this before every launch.