How much a test actually costs to run properly, worked from your target cost per result rather than a number somebody quoted on a podcast. Free, no signup.
Daily, per ad set
$250
Total for the test
$5,250
3 ad sets over 7 days
Spend per creative
$438
The maths: 50 results per ad set at $35 each, because an ad set needs roughly 50 optimisation events a week to leave the learning phase. Fewer events than that and the result is noise rather than a decision.
Budget set, now build the test in one publish
Once you know what the test costs, the slow part is building it. Volume Creatives turns your creatives, copy and ad sets into every combination and publishes the batch paused, named and tagged.
Enter your target cost per result
Use the cost per optimisation event you can afford, not the one you hope for. For purchase campaigns that is your max CPA from the ROAS calculator. For lead or add to cart campaigns, use your recent 30 day average from Ads Manager.
Set the number of ad sets
Count every ad set that will run at once in the test. Each one has its own learning phase and needs its own 50 events, so the total scales in a straight line with this number.
Choose days to decision
How long you intend to run before judging. Seven days is the reference window Meta uses. Shorter windows need a higher daily budget to buy the same events, longer windows need more total.
Set creatives per ad set and read the result
The tool shows daily spend per ad set, the total for the whole test, and how much each creative gets. If per creative spend falls below three times your target cost, it warns you that individual creatives will not get a readable result.
You enter
You get
50 events × $35 ÷ 7 days
$1,750 per ad set × 4 ad sets
$1,750 ÷ 5 creatives, about 10 results each
Four ad sets at $250 a day is a $7,000 week. If that is more than the account can commit, the right move is two ad sets at $250, not four at $125. Halving the daily budget halves the events, and 25 events in seven days leaves every ad set in learning with nothing to decide on.
| Target cost per result | Daily per ad set | 7 days, one ad set | 7 days, three ad sets |
|---|---|---|---|
| $10 | $71 | $500 | $1,500 |
| $20 | $143 | $1,000 | $3,000 |
| $35 | $250 | $1,750 | $5,250 |
| $50 | $357 | $2,500 | $7,500 |
| $80 | $571 | $4,000 | $12,000 |
| $120 | $857 | $6,000 | $18,000 |
Daily per ad set = target cost × 50 ÷ 7. Weekly per ad set = target cost × 50. The last column shows a typical three ad set test. Round figures; the tool computes exact values.
The honest answer is a calculation, not a number. An ad set needs roughly 50 optimisation events in a rolling seven day window to leave Meta's learning phase. Until it does, delivery is unstable and the results are noisy enough that acting on them is guesswork. So the budget a test needs is simply the cost of buying those events: 50 results priced at your target cost per result, per ad set, over the window you intend to run.
Events needed per ad set = 50 × (days ÷ 7). Budget per ad set = events × target cost per result. Daily budget per ad set = that figure ÷ days, which simplifies to target cost × 50 ÷ 7, or about 7.14 times your target cost every day. Total test budget = budget per ad set × number of ad sets. Spend per creative = budget per ad set ÷ creatives in it. A seven day test always needs exactly 50 events per ad set; a 14 day test needs 100, because the window is rolling and the ad set has to keep the pace up, not just hit 50 once.
Every ad set carries its own learning requirement. Splitting a fixed budget across six ad sets does not give you six reads, it gives you six underfed ad sets that all stay in learning and none of which produce a decision. If the calculator says your budget only supports two ad sets, run two. Structural discipline beats thoroughness that cannot be funded. If you genuinely need to test six audiences on a small budget, run them in sequence, two at a time, rather than all at once. See how many ad sets a campaign should have for the structural side of that decision.
The same logic applies one level down. Budget flows to whichever creative Meta favours early, so adding more creatives to a thin ad set mostly produces variants with almost no data attached. As a working rule, each creative wants enough spend to produce at least a few results, which is why the calculator warns when the per creative spend drops under three times your target cost. In practice the split is far from even. It is common for one ad in a set of five to take 60 to 70% of the spend within two days, leaving the other four with a handful of impressions each. Use the combination calculator to see how quickly a test matrix grows, and the ROAS calculator to sanity check the target itself.
This input does most of the work, so it deserves care. For a purchase campaign, the ceiling is your max CPA, which is average order value times gross margin. Planning a test at a cost you cannot afford wastes the test even if it exits learning. For upper funnel events (add to cart, lead, landing page view) use the trailing 30 day average from Ads Manager, not the best week you ever had. If you have no history, take a realistic figure for your vertical and add 30%, since new accounts and new creatives almost always start above their eventual cost. The good CPA guide covers how that number shifts by vertical.
It assumes your target cost per result is roughly achievable. If your real cost lands at double the target, the test needs double the budget or double the time. It also ignores the first day or two of volatile delivery, so treat the total as a floor rather than a precise forecast. It does not model campaign budget optimisation, where Meta shifts money between ad sets and some never reach their 50; if you test on CBO, budget as though the weakest ad set gets a third less than its share. And it assumes 50 is the threshold, which is Meta's published reference rather than a guarantee. Some accounts exit sooner, some sit in learning limited at 60. The point is not exactness. It is to stop tests being launched that could never have reached a conclusion. Use the learning phase calculator to check a single ad set's daily budget against the threshold once the test is live.
Enough to produce roughly 50 optimisation events in seven days, which works out to about seven results a day. Multiply your target cost per result by 50 and divide by 7. At a $35 CPA that is $250 a day per ad set. Below that pace the ad set can sit in learning limited and results stay noisy.
Meta's delivery system exits the learning phase once an ad set has accumulated about 50 optimisation events within a seven day window. That number comes from Meta's own documentation and is the minimum data the algorithm needs to stabilise delivery. The calculator prices those 50 events at your target cost to get the budget per ad set.
Yes, but be clear about what you are buying. Under the threshold you will still get some signal, but delivery stays volatile and cost per result usually runs higher, so a creative that looks bad might just be under fed. The safer way to cut budget is to cut ad sets or creatives, not to spread the same money thinner.
Not in this calculator, because Meta distributes an ad set's budget across its ads rather than funding each ad separately. What creatives change is how far that budget stretches. With four creatives in a $1,750 ad set each gets about $437 on average, but in practice Meta skews spend toward whichever ad wins early, so the others get far less.
Seven days is the usual minimum because it matches Meta's learning window and covers a full weekly cycle of buyer behaviour. Running fewer days forces a higher daily budget to buy the same 50 events and often trips Meta's spend pacing. Running longer than 14 days on the same creatives risks creative fatigue muddying the read.
For a clean creative test, ad set budgets (ABO) give every ad set the amount you calculated here. Campaign budget optimisation (CBO) lets Meta shift money toward whichever ad set it favours first, which is good for scaling but means some ad sets in a test never reach 50 events. Test on ABO, scale on CBO.
Then the test needs proportionally more budget or more time. If you planned at $35 and the first three days average $50, you are buying events 30% slower and a seven day test becomes a ten day test at the same spend. Check cost per result at day three and adjust the window rather than killing the test early.
Connect your ad account, drop in the creatives, and watch the batch land paused. Nothing spends until you say so.