At EUR 30 a day you can resolve about seven ad variations over two weeks, and about three in a single week. Twelve is not a set size at that budget, it is twelve ads each getting under 3,000 impressions, which is not enough to tell a winner from noise. The number is not a matter of taste: it comes out of dividing your test budget by the cost of one reliable read, and if you run more than that you pay the same money for less information.
Small budgets are not a reason to skip variation testing. They are a reason to run a smaller set for longer and to be strict about what you measure.
The largest set a small budget can actually read
One read is the impression count a variation needs before a difference in hook rate or CTR is usable. On Meta feed and Reels, use 5,000 impressions as a planning read, 3,000 if you only want direction, 8,000 if you expect small differences.
largest resolvable set = floor(test budget / (read impressions × CPM / 1000))
Worked, at EUR 30 a day and a EUR 12 CPM. Two weeks of budget is EUR 420. One 5,000-impression read costs 5,000 × 12 / 1000, which is EUR 60. So EUR 420 divided by EUR 60 is 7 variations.
Run twelve instead and the arithmetic goes the other way. EUR 420 at a EUR 12 CPM buys about 35,000 impressions, which is 2,917 each across twelve. At the 1.62 percent Meta feed video CTR in our 2026 ad fatigue benchmark, a synthesis of published figures rather than our own measurement, that is roughly 47 clicks per variation. Nobody makes a buying decision on 47 clicks.
| CPM | Cost of one 5,000-impression read | Resolvable set at EUR 15/day | At EUR 30/day | At EUR 50/day |
|---|---|---|---|---|
| EUR 8 | EUR 40 | 5 | 10 | 17 |
| EUR 12 | EUR 60 | 3 | 7 | 11 |
| EUR 20 | EUR 100 | 2 | 4 | 7 |
| EUR 30 | EUR 150 | 1 | 2 | 4 |
All figures are over a 14-day window. Use your own CPM rather than ours: the gap between a EUR 8 and a EUR 20 CPM changes the answer by a factor of two and a half. Our creative testing calculator runs the same division on your inputs.
Notice the top-right corner. The 8 to 20 live variations our fatigue benchmark puts on a typical active campaign is a real target, and reaching the bottom of it at a EUR 12 CPM takes roughly EUR 480 a fortnight, or about EUR 35 a day. Below that you are not running a smaller version of the recommended set. You are running a different test.
A two-week ladder
Three live at a time, seven-day rounds, one decision per round. This is what fits EUR 30 a day at a EUR 12 CPM.
| Round | Live variations | Budget | Impressions each | Decision |
|---|---|---|---|---|
| Days 1 to 7 | 3 new | EUR 210 | about 5,800 | Kill the lowest hook rate, keep two |
| Days 8 to 14 | 1 winner plus 2 new | EUR 210 | about 5,800 | Kill the lowest, keep the best two |
| Days 15 to 21 | 2 survivors plus 2 new | EUR 210 | about 5,800 | Retire anything past three weeks live |
| Days 22 onward | Steady 3 to 4 live, 2 new per week | EUR 210 per week | about 4,400 to 5,800 | Raise the set only when budget rises |
Decide on hook rate, not CTR or CPA. Hook rate accumulates in the first three seconds of every impression, so it reaches a usable sample days before conversion data does, and at this budget conversion data may never arrive. Typical hook rates run 15 to 25 percent on Meta feed and 20 to 30 percent on TikTok in-feed, so a variation sitting at 9 percent is losing and you need no statistics to say so.
The retirement rule in round three is not optional. Our fatigue benchmark puts week three at 45 to 70 percent below launch baseline and most creative as effectively dead within three weeks, so a variation still running in week four is dragging the campaign average down while you defend it.
To grow the set, work backwards through the same formula. Each extra EUR 15 a day at a EUR 12 CPM buys 15 × 7 divided by 60, which is 1.75 more readable slots per week. Going from a set of 3 to a set of 8 therefore needs roughly EUR 43 a day more, not a new tool. Budget is the constraint at this size, not creative supply.
Why the free plan is genuinely enough to start
Genyad is our product, so this is disclosure rather than a recommendation from nowhere. The free plan gives 5 video ad variations plus one AI-generated video of up to 10 seconds, with no card required and no watermark on any plan.
Count what the ladder consumes. Round one uses 3 variations, round two adds 2. That is 5, exactly the free allocation, so the first fortnight of a small-budget test costs nothing in tooling. Credits never expire, and editing and re-exporting cost nothing, so the 5 do not decay while you think.
Being straight about the limit: 5 is a total, not a monthly refill. From round three you need 2 new variations a week, so 2 credits a week, about 9 a month. The €29 Starter pack is 15 credits at €1.93 each and covers roughly seven weeks of that cadence. The €99 Growth pack at €1.52 a credit only pays off at a higher weekly cadence than this media budget supports. Full detail is on our pricing page, and there is a longer write-up of what the free video ad tool actually does.
What we do not have, so you can rule it out before signing up: no AI avatars or synthetic presenters, no static banner formats, no product-URL import, no product-feed or CSV-driven template rendering, no predicted performance scores, and no direct publishing into Meta or TikTok. You download exports at 9:16, 4:5, 1:1 and 16:9 in 1080p and upload them yourself. With no usable footage at all we are the wrong tool, since every variation comes from a library you upload.
What to skip entirely at this budget
- Dynamic creative permutations. One video against six headlines produces combination-level data that a 35,000-impression fortnight cannot read. Test three videos with one strong caption each.
- Audience testing in parallel. Change one thing per round. Splitting a small budget across creative and audience gives you two unreadable tests instead of one readable one.
- Every export ratio for placements you do not buy. Ship 9:16 and 4:5, add 1:1 and 16:9 later.
- Statistical significance as a gate. You will not reach it at this spend, and waiting for it means never deciding. Rank on hook rate and accept that you are making judgement calls.
- Monthly subscriptions with a floor. Creatify lists $39 a month and Arcads lists €100 a month for 10 videos, both as publicly listed in August 2026, and these prices move. Against a media budget of about EUR 900 a month, a tool at the higher end takes more than a tenth of everything you spend before a single impression is served. Credit packs that do not expire fit an irregular cadence better, which is why ours work that way.
- Weekly refresh on Meta feed. Fortnightly is enough there. Weekly belongs on TikTok and app install, where decay is faster.
Frequently asked questions
How many ad variations should I run on EUR 30 a day?
Three at a time, refreshed weekly, which resolves about seven over a two-week window at a EUR 12 CPM. The calculation is your two-week budget divided by the cost of one read, where a read is 5,000 impressions on Meta and costs 5,000 × CPM / 1000. Check your own CPM first, because a EUR 20 CPM cuts that set from seven to four.
Can I test twelve variations on a small budget if I run them for longer?
Not usefully, because the whole set decays while you wait. Twelve reads at EUR 60 each is EUR 720, which takes 24 days at EUR 30 a day, and our 2026 fatigue benchmark puts week three at 45 to 70 percent below launch baseline. The set is dying before any of it has been read. Fewer variations refreshed more often beats a large set read slowly.
Is a free video ad tool good enough for a real campaign?
For the first two weeks of a three-variation ladder, yes: five variations is exactly what two rounds of that test consumes, and our free plan exports at 1080p with no watermark. It stops being enough at round three, when you need two new variations a week and the free allocation is spent. That is a real limit rather than a hidden one, and it costs about the price of a small credit pack to pass.
What should I measure when the budget is too small for CPA data?
Hook rate first, then CTR, then cost per click, with CPA as confirmation rather than the deciding metric. Hook rate accumulates on every impression, so it is usable within days, while conversion data at EUR 30 a day can stay too thin to rank variations for weeks. Kill on hook rate, keep on CTR, argue about CPA only once a variation has survived two rounds.