DECISIONS, EXPLAINED
See the thinking behind the doing.
Every threshold. Every action. A starting point you can adapt to your economics.
| When | Condition | Action | Defaults |
|---|---|---|---|
| Daily Test BatchDaily | Headroom under capTests scale with free budget, not with ambition | 1 test + 1 per $2,500 headroom, max 8 | 2500 $8 countEditable in profile |
| Creative Loser ControlClosed cohort | ROAS d5 < target on ≥ $300 closed spendHalf of early 'expensive' flags ended at 106–111% ROAS | Pause | 300 $Editable in profile |
| Test AutopilotAny day | $150 spent, 0 purchasesEnough spend to expect at least one purchase | Pause | 150 $Editable in profile |
| Test AutopilotDay 1 | CPT > 1.7 × breakevenDay 1 is noisy; only very expensive ad sets die | Pause | 1.7 xEditable in profile |
| Test AutopilotDays 3 and 5 | 3-day CPT > 1.3 × breakevenEnough data to trust the price | Pause | 1.3 xEditable in profile |
| Test AutopilotDay 7+ | ≥ 20 closed trials and ROAS < 0.7 × targetClosed cohorts confirm the loss | Pause | 20 count0.7 xEditable in profile |
| Test AutopilotDay 14+ | ≥ 30 trials and ROAS < 0.9 × targetNear-target needs more evidence | Pause | 30 count0.9 xEditable in profile |
| Test AutopilotDay 1 | CPT ≤ 0.6 × be, ≥ 7 purchases, spend ≥ 80% budgetStrong early winners shouldn't wait | Budget × 2 | 0.6 x7 countEditable in profile |
| Test AutopilotDay 3+ | CPT ≤ 0.7 × beScale by marginal return | Budget × 3 | 0.7 xEditable in profile |
| Test AutopilotDay 3+ | CPT ≤ 0.82 × be | Budget × 2 | 0.82 xEditable in profile |
| Test AutopilotDay after raise | CPT > 1.2 × beRaises that break efficiency are undone fast | Previous budget | 1.2 xEditable in profile |
| Test AutopilotAny | CPT > be two days, or one day > 1.6 × be | Base budget | 1.6 xEditable in profile |
| Test AutopilotAny | — | Max $2,000 per ad set; proven up to $5,000 in +25% steps | 2000 $5000 $Editable in profile |
| Test AutopilotAny | Near targetCut, don't kill, almost-profitable ad sets | −30%, never below $300 | 30 %300 $Editable in profile |
| Test AutopilotAny | ≥ 3 zero-purchase kills AND > 50% of live ad setsMass failure is usually tracking, not creatives | Kill nothing, alert | 50 %Editable in profile |
| Daily Recommendations + 1-Click ApplyAny | Yesterday's CPT > 5 × median, or data > 2 days staleBad data must never drive a scale | Hold | 5 xEditable in profile |
| Daily Recommendations + 1-Click ApplyDaily | CPT ≤ 0.85 × target CPT | Budget × 2 (under $2k) | 0.85 xEditable in profile |
| Daily Recommendations + 1-Click ApplyDaily | CPT > 1.5 × target CPT | Switch off | 1.5 xEditable in profile |
| Daily Recommendations + 1-Click ApplyBudget ≥ $2k | One bad dayBig winners get trims, not kills | −10%, not kill | 10 %Editable in profile |
| Portfolio Spend GovernorDaily | Share of capSlow new tests before cutting winners | <92% normal · 92–96% free room · 96–100% swap · ≥100% half | 92 %96 %Editable in profile |
Catalog defaults are examples to review, not universal recommendations. Validate against your own targets and data.