The full public record of LaunchDeck optimizer decisions — timestamp, what shifted, and why. Updated every cycle. Pair every entry with the running 30% per-cycle cap and you've got the transparency layer people keep asking for.
+20% into Max Conversions — High Intent Placements
Placements flagged as high intent by the performance model held 2.9 ROAS at $52 CPA over 72h. Reallocated within the 30% per-cycle shift cap; low-intent placements still running as holdout.
+20.0%
-15% out of Prospecting — Interest Layer 3
Layer 3 dropped to 0.7 ROAS over a 5-day rolling window while CPA climbed to $134. Reduced exposure; Layer 1 and Layer 2 absorb the freed budget.
-15.0%
+9% from Brand Search — Generic to Brand Search — Exact
Exact-match holdout lifted incremental conversions 14% over 14 days vs generic. Moved 9% of daily budget to tighten match types without eliminating generic reach.
+9.0%
+25% into Q3 Launch — Creative B
Variant B held a 2.6 ROAS at $48 CPA across 4 days of stable spend, versus A at 1.4 ROAS / $89 CPA. Bumped share-of-voice on B; A remains running for further signal.
+25.0%
-12% out of Retargeting — Site Visitors
Audience saturation: frequency 3.8 and CPM up 22% week over week. Trimmed budget and seeded into Brand Search — Exact Match which held a steady 4.1 ROAS.
-12.0%
+18% from Spring Promo — Broad to Spring Promo — LAL
Lookalike audience trended from 1.4 to 2.1 ROAS over a 7-day window while broad fatigue decayed CTR from 2.3% to 1.1%. Reallocation stays within the 30% per-cycle shift cap.