A creative test has two price tags: the media spend needed to resolve it, and the production spend needed to fill it. At a EUR 12 CPM, a hook-rate read costs roughly EUR 24 of media per variation, and the same variation costs anywhere from a couple of euros to a few hundred to make. Budget one and not the other and you get the same failure every time: a test plan the team cannot supply, or a pile of finished ads with no money left to read them.
Most planning documents we see get the media half right to two decimal places and treat production as free, because the editor is already on payroll. That editor is a capacity limit with a euro value, and once you put the number in, the whole test design changes.
How much media does one variation need?
Media cost per variation is set by the question you are asking, not by the ad. A hook-rate comparison resolves fast because hook rate is a high-frequency event: roughly 2,000 impressions per variation is enough to separate a 30 percent hook from an 18 percent one. At a EUR 12 CPM that is about EUR 24 per variation.
Cost per result is a different animal. You need something like 15 conversions per variation before a cost-per-result difference means anything, and conversions are rare events sitting at the far end of the funnel. That is one to two orders of magnitude more spend than the hook read.
| Question you are asking | Data each variation needs | Media per variation at a EUR 12 CPM |
|---|---|---|
| Which hook holds attention? | about 2,000 impressions | about EUR 24 |
| Which variation earns more clicks? | 5,000 impressions is a fair planning read | about EUR 60 |
| Which variation converts cheaper? | about 15 conversions | EUR 300 at a EUR 20 cost per result, EUR 900 at EUR 60 |
| Which of two audiences responds better? | both of the above, per audience | double the relevant row |
The practical consequence: run the cheap question first. Screen ten variations on hook rate for around EUR 240, kill the bottom half, and only spend conversion money on the survivors. Teams that skip the screen and send twelve variations straight into a cost-per-result test are usually asking for a read they have not funded. You can sanity-check your own numbers with our creative testing calculator, which turns a budget and a CPM into the number of variations it can actually resolve.
How much does a variation cost to produce?
This is the line that goes missing. Three honest reference points, and they differ by a factor of a hundred.
| Who makes it | Effort or fee per ad | Cost per variation | 40 variations |
|---|---|---|---|
| In-house editor | 30 to 90 minutes at a EUR 45 loaded hour | EUR 22 to EUR 68 | EUR 900 to EUR 2,700 |
| Agency or production partner | fixed fee per deliverable | EUR 100 to EUR 400 | EUR 4,000 to EUR 16,000 |
| Genyad, our own tool | 1 credit per variation | about EUR 1.52 on the EUR 99 pack | 40 credits |
The in-house number is the one people argue about. Thirty minutes is realistic when you are swapping a hook onto an existing timeline. Ninety is realistic when someone has to find the right shot, rewrite the voiceover, reset the captions and export four ratios. Multiply by the loaded hourly cost of the person doing it, not their salary divided by 2,080, and the figure stops looking free.
Agencies at EUR 100 to EUR 400 per ad are not being greedy, and for a hero film that number is the right buy. It is simply the wrong instrument for a hook screen, because you cannot afford to throw away eight of ten deliverables at that price, which means you will not throw them away, which means the test stops being a test.
Genyad is our product, so read this with that in mind. It works on footage you already own: you upload the library once, it transcribes and tags every clip, then each variation is a fresh script, shot selection, voiceover, caption set and export rather than a re-cut of one timeline. Pricing is credit-based with no subscription, and the Genyad pricing page lists the packs. What it does not do matters for your budget: no AI avatars or synthetic presenters, no static banner formats, no product-URL import, no product-feed or CSV template rendering, no predicted performance scores, and no direct publishing into Meta or TikTok. Somebody on your team still uploads and names the ads, so keep a line for that.
Cost per published ad, once you apply the keep rate
Cost per variation is a vanity number. The number that governs your budget is cost per published ad, which is cost per variation divided by your keep rate: published divided by produced.
Almost nobody publishes everything they make. A team with a strong brief and one reviewer might keep 80 percent. A team exploring new angles keeps 20 to 30 percent, and that is a sign of ambition rather than sloppiness. Here is what ten published ads cost at each keep rate.
| Keep rate | Produced for 10 published | In-house at EUR 45 each | Agency at EUR 250 each | Credit tool at EUR 1.52 each |
|---|---|---|---|---|
| 80 percent | 13 | EUR 585 | EUR 3,250 | EUR 20 |
| 50 percent | 20 | EUR 900 | EUR 5,000 | EUR 30 |
| 30 percent | 34 | EUR 1,530 | EUR 8,500 | EUR 52 |
| 20 percent | 50 | EUR 2,250 | EUR 12,500 | EUR 76 |
Notice the direction of the effect. A falling keep rate is a rounding error on a credit-priced tool and a budget crisis on an agency retainer. That is the real reason agency-supplied accounts test conservatively: the cost of a rejected concept is high enough to change behaviour, and the behaviour it changes is exactly the exploratory testing that finds new winners.
A worked monthly figure
Take a target of 20 published variations a month, which sits inside the 15 to 50 range our 2026 ad fatigue benchmark associates with 3 to 5 times longer campaign lifespan than quarterly refreshes. At a 50 percent keep rate that is 40 produced.
Media, in two stages. Hook-screen all 40 at 2,000 impressions each: 80,000 impressions, EUR 960 at a EUR 12 CPM. Promote the best 8 to a conversion read at 15 conversions each and a EUR 20 cost per result: EUR 2,400. Media total for the month, about EUR 3,360.
Production for the same 40: EUR 900 to EUR 2,700 in-house, EUR 4,000 to EUR 16,000 through an agency, or 40 credits on a credit-based tool, which is a little over EUR 60 at the Growth rate.
So the production line runs from under 2 percent of the media budget to nearly five times it, depending only on who makes the ads. That spread is why the production question deserves the same scrutiny as the CPM. Our video ad cost calculator will run your own hourly rate and keep rate through the same arithmetic.
Where these budgets usually break
Three failures, in order of how often we see them.
The team funds the media and staffs the production at half the required rate, so the test launches with 6 variations instead of 14 and never reaches a clean read. The team funds both but forgets that fatigue keeps consuming inventory: the same benchmark puts week three at 45 to 70 percent below launch baseline, so a monthly budget with no replacement stock built in is a budget for two good weeks. And the team counts production only for winners, which makes the cost per published ad look four times better than it is.
Frequently asked questions
How much should I spend on creative testing each month?
Work it backwards from the number of published variations you need, not from a percentage of total spend. Twenty published variations a month at a 50 percent keep rate means 40 produced, roughly EUR 3,400 of media at a EUR 12 CPM across a hook screen plus a conversion read on the survivors, and a production line anywhere from EUR 60 to EUR 16,000 depending on who makes them.
What is a reasonable cost per published ad?
Divide your cost per variation by your keep rate and judge that number, not the sticker price. In-house editing at 30 to 90 minutes per ad and a EUR 45 loaded hour lands between about EUR 28 and EUR 340 per published ad across realistic keep rates, agency work at EUR 100 to EUR 400 per deliverable runs several times higher, and credit-based generation from existing footage sits in the low tens of euros.
Does a hook-rate test really need only 2,000 impressions per variation?
Roughly, yes, because hook rate is measured on a very common event and the differences you care about are large. It will tell you which opening holds attention and it will not tell you which ad converts cheaper. Treat it as a screening instrument that qualifies variations for a more expensive test.
Should the creative testing budget sit inside the campaign budget?
Keep the media portion inside the campaign so it competes for the same auctions and inherits the same audiences, but track it as a separate line so nobody quietly reallocates it to the scaling ad set. The production portion belongs in its own budget, because it is a capacity commitment that has to be funded before the month starts rather than adjusted mid-flight.