What a creative testing budget actually buys
A creative test is a purchase. You are buying one piece of information per concept: its real CPA. The budget question is not “what can we afford to test” but “what does a readable answer cost,” and that has a formula.
Total test budget = Budget per concept × Concepts in test
Days to complete = Total test budget ÷ Daily budget
Most creative tests fail before launch, at this step. Teams pick a number of concepts first, split whatever budget exists across them, and end the month with six half-read cells instead of three answers. This calculator runs the math in the honest direction: answers first, then how many you can afford.
Worked example
Take the calculator's defaults: a $62 expected CPA, 50 orders per concept, 4 concepts, and $400 per day of test spend. Pull your own CPA from the last 30 days of account data; the CPA Calculator turns spend and conversions into that number.
$3,100 × 4 = $12,400 total test budget
$12,400 ÷ $400 = 31 days
So this test is a $12,400, month-long commitment. If 31 days is too slow, you have exactly two levers: fewer concepts or more daily budget. At $800 per day the same test finishes in 15.5 days. What you cannot do is keep 4 concepts, keep $400 a day, and expect the answer sooner. The orders arrive when they arrive.
What to plug into each input
Expected CPA: use the CPA of the traffic the test will actually run on. Most creative tests run cold prospecting, where CPA sits above the blended number retargeting flatters. Plug in the prospecting figure, not the account-wide one.
Orders per concept: 50 is the planning floor for a CPA read you can act on. Directionally, coarse kill-or-keep calls can get by nearer 30; thin margins and expensive decisions justify 75 or more.
Concepts in test: count concepts, not variations. A new hook on the same body is a variation: it shares its concept's budget instead of claiming a cell of its own.
Daily test budget: the number you can hold for the entire run without raiding it for scaling mid-test. Defund the default test at day 18 and you have spent $7,200 for four cells sitting near 29 orders each: half-read, decision-free.
Why ~50 orders per concept
CPA is spend divided by orders, and small denominators lie. At 10 orders, one extra conversion moves a concept's CPA by 9%. One lucky three-order day moves it by 23%, enough to flip your ranking of two concepts overnight. At 50 orders, that same lucky day moves CPA by under 6%, and the ranking mostly stops flapping.
Treat 50 as a directional planning floor, not a statistical guarantee. It is the point where acting on the read is defensible, not the point where a statistician signs off. If two concepts finish within a few percent of each other, the honest answer is “tied.” If you need a formal verdict with a minimum detectable effect and real power, run the numbers through the A/B Test Sample Size Calculator. It will usually ask for more than 50 orders. That is the price of “proven” over “probable.”
Kill losers early, read winners fully
The 50-order budget is what a full read costs, but only potential winners deserve a full read. Losers usually announce themselves far upstream, in metrics that accumulate hundreds of times faster than orders: hook rate shows up within a day of impressions, CPC within a few hundred clicks.
If a video's hook rate runs at half your account norm, spending it to 50 orders is buying information you already own. Kill it, and its unspent budget flows back to the concepts still standing. The asymmetry matters: proxies are good enough to kill, never good enough to crown. A scroll-stopping hook that doesn't convert is one of the most common patterns in paid social, which is exactly why winners must earn their verdict in orders, not in hook rate alone.
Testing cadence vs. budget reality
Here is the collision most teams don't see coming: $400 a day is roughly $12,000 a month, and the default test above costs $12,400. One 4-concept test cycle owns the entire month. Want 8 fresh concepts a month at a $62 CPA? That is $24,800 of testing spend, before a dollar goes to scaling what wins.
So your testing cadence has a hard budget ceiling: concepts per month = monthly test budget ÷ (expected CPA × orders per concept). Run your real numbers through the calculator and the ceiling stops being a surprise. Teams below the ceiling are constrained by spend. Teams at it are constrained by creative production, a different problem with a different fix.
Frequently asked questions
Why 50 orders per concept?
Because CPA reads are noisy at small order counts: at 10 orders a single conversion moves the number 9%, so a lucky day can reorder your whole test. Around 50 orders, day-to-day noise shrinks to a few percent and the ranking stabilizes. It is a directional floor for media buying decisions, not a significance threshold; the sample size calculator handles the formal version.
Can I judge creatives on CPC or hook rate instead?
You can kill with them, not crown with them. Hook rate and CPC accumulate fast, so a concept that badly trails the account on either is safe to cut early. But strong-hook, weak-conversion creatives are common: a proxy winner is a hypothesis, not a result. Promotion to scaling should always be paid for in orders.
How many concepts should I test at once?
As many as your budget can carry to a full read inside about a month. The calculator makes that trade explicit. Three concepts fully read beat six concepts half-read, because half-read cells produce decisions you quietly reverse later. At typical DTC budgets that lands at 2-4 concepts per cycle, which is directional, not doctrine.
When is it OK to kill a creative early?
When the signal is upstream and the gap is large: hook rate or CTR at roughly half your account norm, or CPC far above it, with a day or two of real delivery behind the numbers. A spend-based tripwire also works: for example, 2-3× your expected CPA spent with zero orders is a fair early exit. Small gaps deserve patience; big upstream gaps don't.
My test budget is bigger than my scaling budget. Is that wrong?
Not automatically. Young accounts spend most of their money learning, and that is the fastest route to a winner worth scaling. It becomes a problem when it persists: testing exists to feed scaling, so if months pass and nothing graduates, the bottleneck is creative quality, not budget allocation. Fix the inputs to the test before shrinking it.
What CPA should I plug in?
Start with your last-30-day account CPA; if the test runs cold prospecting only, use prospecting CPA, since retargeting flatters the blend. New concepts usually start worse than the account average because they launch without optimization history, so if anything, round up. Plugging in your best-ever campaign CPA sizes a test that runs out of money before it runs out of questions.
Where StefanBrain fits
This calculator prices the test; someone still has to produce 4 concepts a month worth spending $3,100 each to read. StefanBrain generates the static ads, video ads, and copy variations that fill those testing slots from your brand, offers, and past winners, so cadence becomes the budget math above, not a production bottleneck. Then sanity-check the whole plan with the Ad Budget Calculator to make sure the testing line and the revenue goal fit inside the same media budget.