Testing and scalingGuide
How many ad creatives? Plan useful variety first
Short answer
There is no universal creative count that makes an ad set useful. Plan distinct ideas, then fit the variants to the team’s production and review capacity and the tests that can be interpreted. Separate new concepts from small iterations, record the hypothesis behind each change, and avoid treating more files as more learning.
Ten exports can still repeat one argument. Plan creative variety around differences a reviewer can name: audience situation, evidence, opening action, or explanation. Record what each variation is meant to explore before production, then keep its relationship to the source material clear so file count does not stand in for a testing decision. The broader creative fatigue guide explains why a new production decision may be needed after repeated exposure.
Define what makes two ideas different
Compare the viewer problem, the explanation, and the evidence. If those stay the same while a caption color changes, the result is an iteration of one idea. That may be useful, but it should not be counted as a wholly new concept.
In our production in October 2026, scripts across a batch should not share interchangeable hooks, turns, and endings. Apply that rule by reading the ideas without their format names. If two scripts can swap their middle sentences unnoticed, identify what new question the second one answers before making more versions of it.
Separate concepts from iterations
Keep a concept ID for the underlying idea and a version ID for the change. An opening revision, a language version, and a new product argument are different kinds of work. The naming should make that distinction visible to the reviewer.
A hypothetical storage-product batch might contain one idea about finding an item, another about a particular opening action, and a third about arranging a space. A new voice on the first idea remains a voice iteration. That keeps production variety honest without dismissing the practical value of a targeted refinement.
| Change | Suggested classification |
|---|---|
| New product argument | New concept candidate |
| Same body, different opening | Opening iteration |
| Approved message in another language | Language version |
| Same edit, changed caption treatment | Presentation iteration |
Source: suggested production checklist, October 2026.
Plan only what the team can review
Count the actual work each candidate needs: script approval, source selection, voice, edit, and final inspection. A batch that outruns review capacity can accumulate unapproved files rather than usable ads.
Choose the next set from the questions the team can answer. A small team may need to finish one distinct message before making several superficial variants. This is a planning recommendation, not a fixed count for every advertiser. The appropriate batch depends on the available sources, the owner’s decisions, and the tests the team can interpret responsibly.
Hypothetical bottle concepts are proposed creative differences, not reported test results. One starts with a fit demonstration, while another explains the cap mechanism. Their opening evidence and arguments differ, so each can receive a clear hypothesis in the production record. A caption-color change has a narrower purpose and should be labeled accordingly. This distinction helps a team describe what it made without assuming that three files represent three equally different ideas.
Give each change a testable question
Write the expected observation before producing the version. For an opening iteration, the question might concern whether the object is identified more clearly. Keep the rest of the approved edit stable where that helps isolate the change.
Keep an ad test hypothesis log you can act on covers the hypothesis log. Record what changed and what stayed fixed. A new concept may deliberately change several connected elements because it tells a different story; label it as such rather than pretending the result isolates a single caption or shot. The test description should match the scope of the production change.
Use platform research without inventing a recipe
Meta Andromeda: what creative variety changes explains the documented scope of Meta’s Andromeda retrieval system. Technical platform research can inform why distinct inputs deserve attention, but it does not provide a universal production count or promise delivery to a new audience.
Keep the creative plan grounded in the brand’s evidence and actual messages. Do not manufacture irrelevant versions merely to appear diverse. A clearly different explanation supported by appropriate footage is more useful editorially than many copies whose only difference is a background color. State that as a production judgment, not an asserted platform rule.
- 01Map distinct ideasMessage and evidence
- 02Label iterationsOpening, language, presentation
- 03Write each hypothesisScope of the changed factor
- 04Fit review capacityProduce the next useful batch
Reserve sources for distinct roles
In our production in October 2026, clip reuse is tracked across ads and candidates are rotated instead of always selecting the same first match. A shared usage record makes repetition visible during planning.
Inspect whether multiple ideas rely on the same opening source. If they do, decide whether the repeated image is intentional brand continuity or an avoidable constraint. Do not replace strong product evidence merely to make every frame different. The useful distinction is between necessary repeated proof and habitual reuse that makes separate ideas appear nearly identical.
- File varietyDifferent exports or colorsCreative varietyDifferent arguments or evidence roles
- File varietyCan repeat one openingCreative varietyMakes distinct openings visible
- File varietyMeasures output quantityCreative varietySupports a clear production question
Keep language and place versions in context
City and language versions of one approved ad covers city and language versions built around an approved body. Those versions can serve practical production needs without creating a new concept each time. Review the changed words, voice timing, and relevant visual context separately.
Avoid confusing market coverage with conceptual variety. The same argument in another language may still be the same argument. Likewise, one language can support several distinct messages. A production matrix with concept and version fields helps a reviewer see both kinds of coverage without turning either into an inflated total.
Decide the next batch from the record
Review the completed candidates, unresolved source gaps, and the decisions the test can support. Choose whether the next work should refine an opening, develop a new message, or collect missing proof. Do not make additional exports merely because a target file count has not been reached.
Turntake’s production workflow describes the production approach. Preserve the concept IDs, changed factors, review status, and source dependencies with the batch. The useful output is a portfolio of understandable ideas and bounded iterations, with enough capacity left to inspect them and learn from the questions they were made to answer.
Who wrote this We made the ads first, for SmartSolar in Denmark, Sweden and Ireland. Then we built Turntake. Where Turntake comes from
FAQ
How many creatives should I put in an ad set?
This guide does not prescribe a universal count. Plan distinct ideas, production capacity, review capacity, and interpretable tests first. A large export total can repeat one concept, while a smaller set may answer several clear creative questions.
Is a new hook a new concept?
Usually treat a new opening on the same approved body as an iteration unless it changes the underlying argument. Keep concept and version IDs separate. That makes the scope of the change visible without dismissing the iteration’s practical value.
Does a new animation style create useful variety?
It creates presentation variety, but may preserve the same message and evidence. Review the scripts without style labels to judge conceptual differences. Use a new visual world when it serves the explanation rather than merely increasing the file count.
What should determine the next batch?
Use the outstanding creative questions, source gaps, and decisions from the current record. The next task may be a refinement, a new argument, or filming missing proof. Match the batch to what the team can produce and inspect well.