small-budget-adslisted
Install: claude install-skill RabbitDesigns22/claude-skills
# Small-budget ad decisions (Meta Ads)
Most ad advice assumes thousands per month and clean A/B tests. Small local businesses run 1-5 USD a day per ad set, one ad set per product, with conversations to WhatsApp as the result. With so little data, the usual rules ("kill anything above target CPA after 3 days") turn off good ads and reset the learning phase for nothing. These rules keep decisions safe.
## 0. Permission
- Look only, by default. Turning ads on or off, editing, duplicating, changing budgets or publishing happens only with the account owner's explicit yes, change by change ("Turn off ad 1, 2 and 3?").
- After any change, read it back in Ads Manager and show it.
## 1. Is there enough data?
- Minimum window: 7 full days since the last significant edit. With fewer days, only flag what is clearly broken and say "too few days".
- Today's data is for spotting breakage (0 impressions, rejection, payment failure), not for turning things off. One bad day is not a trend.
- The **benchmark** of each ad set is its own normal cost per conversation: its last 30 days before the period you judge (if it is new, the period itself). Never another ad set's benchmark, and never only the bad period: if the whole ad set dropped, its current average hides the problem.
- An ad can be called "bad" only after spending at least **2x the benchmark**. Below that, zero conversations means nothing.
- Example: an ad set normally costs 10 per conversation. An ad at 14 is 1.4x: keep it. An ad t