How Long to Wait Before Judging a New Ad Campaign
Here's the answer: at $50/day you need 2-3 weeks before any data means anything. At $200/day you need 7-10 days. At $1,000/day you can often make calls in 3-4 days. The number that matters isn't days, it's conversions. You need roughly 50-100 conversions per ad variation before the data is statistically useful. Work backward from your conversion rate and daily spend to figure out your real timeline.
This confusion is everywhere because the ad platforms train you to check daily. The dashboard refreshes every hour. Facebook sends you a notification when spend hits a milestone. Everything about the interface says "look now, react now." So people do. They see a campaign down 15% on day two and kill it. They see one ad with a slightly better CTR on day one and dump the whole budget into it. None of that is real signal yet. It's noise dressed up as a trend line.
"Give it 24-48 hours and you'll know if it's working"
This one comes from a half-truth. The learning phase on most platforms does need a day or two to exit, and your early impressions will look weird because the algorithm is still figuring out who to show the ad to. So yes, the first 24-48 hours are usually garbage data. But the myth twists this into "48 hours is enough to judge," when really 48 hours is just enough to stop judging the garbage part. You still need days or weeks after that before you have a real sample.
I ran paid media on a $2.2B infrastructure project where a single landing page test needed almost three weeks before the numbers stopped bouncing around. Early on, conversion rate swung between 1.2% and 4.8% day to day. By week three it had settled into a tight band around 2.3%. If I'd made a call on day two, I'd have killed the winning version.
"If one ad is clearly beating the others, that's your winner"
Clearly beating is doing a lot of work in that sentence. A 20% difference in CTR with 40 clicks total is not a signal. It's a coin flip that landed on heads four times out of seven. People see percentage differences and assume percentage differences mean something, regardless of sample size.
Here's the actual math people skip: with small sample sizes, random variance alone can produce big-looking gaps. Two ads with a true identical performance of 2% conversion rate can easily show one at 1.4% and one at 2.8% after only 20-30 conversions each, just from normal statistical noise. You need that 50-100 conversion mark per variant before the gap you're seeing is more likely to be real than random.
"Low budget campaigns just take longer, but the process is the same"
Not quite. Low budget campaigns don't just take longer, they often never get there at all if you're testing too many variables at once. This is the part people miss. If you're spending $30/day and running 5 ad variations, you're splitting an already-small sample five ways. You might wait two months and still not have a usable read on any single ad.
The fix isn't patience, it's math before you launch:
- Figure out your expected conversion rate first, even a rough estimate from past campaigns or industry benchmarks.
- Divide your target sample size (50-100 conversions) by that rate to get the clicks you need.
- Divide clicks needed by your expected CTR to get impressions needed.
- Divide impressions needed by your budget's reach-per-dollar to get days.
If that math says 45 days and you only budgeted 2 weeks for the test, you don't have a test. You have an opinion with a spend limit.
What actually matters instead
Stop asking "how long before optimizing a new ad campaign" as if there's one universal number. Ask "how many conversions do I have on each variant right now." That's the real clock. Everything else, days, weeks, gut feeling, is a proxy for that number and a bad one. On a DTC brand I scaled from $100K to $3M in revenue, we built a simple rule: no creative gets killed or scaled until it hits 50 purchases, full stop, regardless of how many days that took. Some ads hit that in 4 days. Some took 3 weeks. The rule didn't care about calendar time. It cared about sample size. That single change in discipline cut our wasted ad spend on premature kills by a real, measurable amount within the first quarter of using it.
The most common mistake I see, by far, is people comparing a 3-day-old campaign to a 30-day-old campaign and treating the comparison as fair. It isn't. Always compare apples to apples: same time window, same spend level, same number of conversions in the bucket. A new campaign isn't underperforming your evergreen winner, it's just younger. Give it the same runway before you compare the two.
Practical takeaway: before you launch, calculate the conversions-per-variant number you need to hit, translate that into an estimated number of days based on your budget, and write that date down somewhere you'll actually see it. Don't touch the campaign before that date except to fix something broken, like a landing page error or an ad that got disapproved. When that date hits, look at conversions, not clicks, not CTR, not vibes. That's the whole method.