Creative testing budget calculator
A Meta creative test costs your CPA times the conversions each variant needs, times the number of variants. Divide that by your daily test budget for the days it takes. Enter your own numbers below to see the budget, the days and the smallest CPA gap the test can reliably detect. Free, no signup.
Every result is a planning estimate for your own account, built from the numbers you enter.
Budget and days from your CPA
Use the cost per result of the conversion you will judge the test on. The variant count and daily budget also feed the CTR and conversion rate plan further down.
Spend per variant
$1,250
CPA × conversions per variant
$25 × 50
Total test budget
$2,500
spend per variant × variants
$1,250 × 2
Days at your daily budget
13 days
total test budget ÷ daily budget, rounded up
$2,500 ÷ $200 = 12.5
What this plan can and cannot tell you
At a two-sided 5% significance level and 80% power, the settings the budget article uses. A planning estimate for your own account, from the numbers above; for small samples, an exact or simulation-based method is safer than this approximation.
Smallest CPA gap it can detect
39%
(1.96 + 0.8416)² × (1 + r) ÷ (r − 1)² = conversions per variant; gap = 1 − 1 ÷ r
r = 1.64 at 50 conversions per variant
If one variant's real CPA is about 39% or more below the other's, 50 conversions each will show the gap about 4 times in 5. When two variants are truly equal, a false winner still appears about 1 time in 20. A smaller real gap will often end in a tie or swap places as data comes in, so treat a narrow lead at this volume as unproven.
Conversions per variant, 20% gap
283
(1.96 + 0.8416)² × (1 + r) ÷ (r − 1)², r = 1 ÷ (1 − gap)
r = 1.25
To detect a gap of 20%, each variant needs about 283 conversions: $14,150 in total at your CPA, or 71 days at $200 a day.
- If each variant runs in its own ad set, Meta's learning phase looks for about 50 optimization events per ad set in the 7 days after the last significant edit. At your CPA that takes about $179 a day per ad set ($25 × 50 ÷ 7). This plan gives each variant $100 a day, about 28 conversions a week, so those ad sets may stay in learning, where results swing more.
- The days assume the budget splits evenly. When several ads share one ad set, Meta can send most of the spend to one of them, and the variant with the least spend sets the end date.
Budget and days from CTR or conversion rate
The budget article's sample-size formula, for a test judged on a rate. It uses the 2 variants and $200 daily budget set above.
Impressions per variant
42,694
[1.96 × √(2p̄(1 − p̄)) + 0.8416 × √(p1(1 − p1) + p2(1 − p2))]² ÷ (p2 − p1)²
p1 = 1%, p2 = 1.2%
Test budget
$1,281
impressions per variant × variants ÷ 1,000 × CPM
85,388 ÷ 1,000 × $15
Days at your daily budget
7 days
test budget ÷ daily budget, rounded up
$1,281 ÷ $200 = 6.4
p1 is the current rate, p2 the target and p̄ their average. Going from 1% to 1.2% is a relative lift of 20%, which each variant needs about 42,694 impressions to show. A CTR test shows which ad earns the click. Check the winner's conversions before you scale it. For rare events or small samples, an exact or simulation-based method is safer than this approximation.
At 50 conversions per variant, this test can separate CPAs about 39% apart. Variants built on different concepts or angles are more likely to differ by that much than small edits of one ad.
See the creative testing workflowHow the calculator works
Three lines turn a CPA into a plan, and a fourth says how large a difference that plan can detect.
Spend per variant = CPA × conversions per variant
Total test budget = spend per variant × number of variants
Days = total test budget ÷ daily test budget, rounded up
Conversions per variant for a CPA gap = (1.96 + 0.8416)² × (1 + r) ÷ (r − 1)², where r = 1 ÷ (1 − gap)
Learning-phase budget per ad set = CPA × 50 ÷ 7, per day
With the defaults, a $25 CPA and 50 conversions per variant make $25 × 50 = $1,250 per variant. Two variants make $2,500, which a $200 daily budget spends in 12.5 days, so plan on 13. Those 50 conversions per variant can detect a CPA gap of about 39%. A 20% gap needs about 283 conversions per variant, or $14,150 for two variants at the same CPA.
The detection line is the two-proportion sample-size formula from the creative testing budget article, at a two-sided 5% significance level (z = 1.96) and 80% power (z = 0.8416). It takes the inputs the NIST/SEMATECH handbook lists for proportion sample sizes: a baseline rate, the difference to detect, the significance level and the power. When each impression rarely converts, the formula depends only on the number of conversions, which is why the CPA view can work from conversions alone. The learning-phase line uses the guidance this site follows: about 50 optimization events per ad set in the 7 days after the last significant edit.
Worked examples from the budget article
Each row is a worked model with two variants, computed by the same formula as the CTR and conversion rate plan above. None of them is a forecast or an AdRiseLab customer result.
| Test | Per variant | Budget, two variants |
|---|---|---|
| Link CTR, 1.0% to 1.2%, $15 CPM | 42,694 impressions | $1,281 |
| Link CTR, 2.0% to 2.4%, $15 CPM | 21,109 impressions | $633 |
| Conversion rate, 3.0% to 3.6%, $2 CPC | 13,914 clicks | $55,656 |
| Conversion rate, 5.0% to 6.0%, $2 CPC | 8,158 clicks | $32,632 |
| Conversion rate, 3.0% to 4.5%, $2 CPC | 2,518 clicks | $10,072 |
The same 20% relative lift costs about $1,281 to detect on link CTR and about $55,656 on conversion rate, because clicks cost far more than impressions. Testing for a wider gap is the cheapest lever: aiming for 3.0% to 4.5% instead of 3.0% to 3.6% cuts the budget from $55,656 to $10,072.
Conversions per variant by CPA gap
How many conversions each variant needs before a CPA gap of a given size shows up reliably, from the formula above at 5% two-sided significance and 80% power.
| CPA gap to detect | Conversions per variant |
|---|---|
| 50% lower CPA | 24 |
| 40% lower CPA | 48 |
| 30% lower CPA | 104 |
| 25% lower CPA | 165 |
| 20% lower CPA | 283 |
| 15% lower CPA | 549 |
| 10% lower CPA | 1,343 |
The requirement climbs fast as the gap narrows: going from a 30% gap to a 15% gap takes it from 104 to 549 conversions per variant.
What the numbers can and cannot tell you
Days are budget arithmetic. A test is done when each variant has the conversions you planned, which takes longer than the days figure when Meta spends unevenly, so judge a test by the conversions it collected.
An equal budget per ad does not make a controlled test by itself. Ads in one ad set share its budget, and delivery can send most of it to one of them. Separate ad sets with their own budgets, or an A/B test from Experiments in Ads Manager, keep each variant's spend closer to the plan.
Meta's learning phase looks for about 50 optimization events per ad set in the 7 days after the last significant edit. That figure is about delivery settling. Whether a test can call a winner depends on the size of the gap, as the table above shows.
Each variant you add costs one more spend per variant and adds days. With more variants, more pairs get compared, so the chance that one looks better by luck rises. One clean comparison answers more than five underpowered ones.
Match the metric to the decision. CTR is cheap to test and shows which ad earns the click. Which ad sells is a conversion question, so check a CTR winner's conversions before you scale it. The Meta ads A/B testing guide covers the setup in Ads Manager, and the creative testing framework covers what to test first.
Frequently asked questions
How much should I budget for a Meta creative test?
Multiply your CPA by the conversions you want each variant to collect, then by the number of variants. At a $25 CPA, 50 conversions per variant and two variants, that is $25 × 50 × 2 = $2,500. Treat it as a planning estimate for your own account: the real spend moves with your CPA while the test runs.
How long should a Meta creative test run?
Until each variant has the conversions you planned. For an estimate, divide the total test budget by the daily test budget: $2,500 at $200 a day is 12.5 days, so about 13. If delivery favors one variant, the variant with the least spend sets the end date. Time alone is not the sample: three days can buy 500 impressions or 500,000.
Are 50 conversions per variant enough to pick a winner?
Only when the gap is large. With 50 conversions each, two variants' CPAs need to be about 39% apart for the test to show it reliably, at 5% two-sided significance and 80% power. A 30% gap needs about 104 conversions per variant, a 20% gap about 283 and a 15% gap about 549. The 50 figure comes from the learning-phase guidance of about 50 optimization events per ad set in 7 days, which describes delivery settling down.
Why is a CTR test so much cheaper than a conversion test?
Because impressions cost far less than clicks. With the budget article's numbers, a 20% lift on a 1.0% link CTR needs 42,694 impressions per variant, about $1,281 for two variants at a $15 CPM. The same 20% lift on a 3.0% conversion rate needs 13,914 clicks per variant, about $55,656 at a $2 CPC. CTR shows which ad earns attention, so check the winner's conversions before you scale it.
Can I test more than two variants with this calculator?
Yes. The budget and days scale with the number of variants, and the detectable CPA gap applies to any one pair of them. With more variants, more pairs get compared, so the chance that one looks better by luck goes up. When the budget is tight, fewer variants with more conversions each give a clearer answer.
Give the test concepts worth comparing
A test budget is worth spending when each variant tests a different idea. Paste a product URL and AdRiseLab builds ad concepts from different hook types and angles, in feed, square and vertical formats. You review every concept, and nothing reaches Meta until you approve.
Cancel any time.