What's the minimum data I need before I trust that a paid campaign is actually working?
Roughly 30 to 50 conversions per variation before you read anything into a result, and enough spend that you're not judging on 3 or 4 lucky sales. Below that, what looks like a winning ad is usually just noise. Give a campaign at least a week or two and a few hundred clicks before you call it dead or alive.
Go deeper
3 resources, 2 link-checked.
▶️ Video
✓ Link checkedFreeBeginner
Before you judge a paid campaign, you need to know the platform spends the first stretch guessing, and this video from a working media buyer breaks that down plainly. It explains why cost per result swings wildly early and why the algorithm needs a batch of conversions (around 50 in a week) before delivery settles. Watch it so you do not kill an ad set that was simply still learning.
When one ad beats another, this free calculator tells you whether the gap is real or just luck. You plug in visitors and conversions for each version and it returns the probability that the difference holds up, no spreadsheet or stats background needed. Use it as a gut check before you shift budget onto the ad that only looks like the winner.
Most founders call a paid campaign a winner or loser off a handful of clicks, and this piece walks through exactly why that read is usually noise. It puts real numbers behind the question: what baseline conversion rate, what minimum detectable effect, and how many conversions you need before the result means anything. Treat it as a starting point for setting your own threshold before you launch, not after you are staring at the dashboard.
