Test creatives without wasting budget
Launch distinct concepts in weekly waves, leave them alone long enough to learn, then cut on cost per sale rather than gut feel.
- A concept is a new angle; an iteration is the same idea with a new hook, visual or format.
- Launch 8–12 genuinely different ads at a time and hold the rest of the batch back.
- No edits for 3–4 days; first real review at day 7.
- Judge an ad once it has spent about 3× your target CPA.
- Add new ads in weekly batches, because each one added is a significant edit.
Concepts and iterations
| What changes | Use it to | |
|---|---|---|
| Concept | The angle or message: problem, outcome, proof, comparison, offer | Explore what works |
| Iteration | The hook, headline, visual or format of a proven idea | Stretch a winner |
Explore with new concepts, then extend the winners with iterations. One agency says at least half the ads in accounts spending over $20k are second- or third-generation iterations of earlier concepts.
How many ads to launch
Motion’s data covers $1.29B in spend across 578,750 creatives. It found that, even at the same budget, brands that launch more creative “get twice the number of winners.” Smaller accounts have lower hit rates.
It also found some of the best-performing formats were among the easiest to make: text-based ads, product images and UGC.
Where to test
There are two schools. Consolidated testing adds new ads to your main campaign and lets the system pick. A separate test campaign gives each batch its own ad set and budget, then moves winners across.
| Approach | Upside | Downside |
|---|---|---|
| Consolidated | Simple, and all the data stays in one place | New ads may get little spend |
| Separate test campaign | Every batch is guaranteed spend | Costs more and splits the data |
For a head-to-head answer, such as “offer card vs testimonial”, use the Creative testing tool in Ads Manager. Rolled out around October 2025, it tests 2–5 ads inside one ad set with equal spend and no audience overlap. Some accounts report a 10-ad limit.
Tests run 7 days by default and can run up to 30. The platform suggests putting no more than 20% of your budget into a test.
The learning phase
The documented rule for leaving the learning phase, as of October 2026.
Significant edits include changing the targeting, the optimization event or the bid strategy, adding a new ad, pausing for more than 7 days, and significant budget changes.
An ad set that can’t reach 50 shows Learning limited. It still delivers, with bigger swings.
When to cut, keep or scale
5× gives a cleaner read. A practitioner heuristic, not a platform rule.
| Conversions at 3× CPA spend | What to do |
|---|---|
| 0–1 | Cut it |
| 2 | Give it about 1× CPA more |
| 3 or more | Keep it |
Heart the ads you’d launch first, then keep the rest for next week’s wave.
Sources (11)
- 1.New Engen, When to iterate and when to diversify, Jul 29, 2026
- 2.Foxwell Digital, How to test creatives in 2026, Jun 24, 2026
- 3.Motion, Creative Benchmarks 2026, Apr 17, 2026
- 4.Motion, How many creatives do you need to launch?, Mar 2026
- 5.Flighted, How to calculate your creative testing budget, Jul 1, 2026
- 6.TheOptimizer, Post-Andromeda testing playbook, May 23, 2026
- 7.Jon Loomer, Creative testing
- 8.Search Engine Journal, Evaluating creative performance, Nov 12, 2025
- 9.MarketingConcurrent, 3 persistent learning-phase myths, Sep 21, 2026
- 10.PPC Land, Learning phase glossary, 2026
- 11.Jon Loomer, Is the learning phase changing?