How do you test ad creatives before spending on media?
TL;DR
- Screen for four things: hook, takeaway, brand linkage, next action.
- Brand linkage is the one teams skip and the one that wastes the most spend: people remember the ad and credit a competitor.
- Split-cable research on 389 TV tests found creative changes more likely to move sales than extra weight alone.
- Neon Blue screens every creative on a simulated audience before the media plan locks, and the winners go to live in-platform tests. Run your first study free
Neon Blue builds simulations of human behavior. You test ad creatives before spending on media by showing each one to people who match your target audience and asking four things: did the opening hold them, what did they take away, which brand was it for, and what would they do next. A creative that fails any of the four rarely gets rescued by budget. Running that screen on a simulated audience lets a team cut a stack of six or ten creatives to the two or three worth paying to test in-platform.
Does the creative matter more than the media weight?
In 1995, Leonard Lodish and colleagues published a meta-analysis of 389 real-world split-cable TV advertising experiments in the Journal of Marketing Research. Households in the same markets saw different ad weights or different copy, and their purchases were tracked. Turning up the weight of an existing campaign often did nothing measurable for sales. Changes in copy, creative and brand strategy were more likely to make a difference. The lesson carries to social and online video: a weak creative with more budget is still a weak creative, so the screening happens before the money moves, the same logic as testing a product idea before launch.
What should a pre-flight creative screen measure?
| Measure | Question | What a failing creative looks like |
|---|---|---|
| Hook | Would you keep watching past the first few seconds? | Most people scroll before the product appears |
| Takeaway | What was this ad trying to tell you? | Answers scatter, or name the wrong benefit |
| Brand linkage | Which brand was this ad for? | People name a rival or "not sure" |
| Next action | What, if anything, would you do after seeing this? | "Nothing" dominates, even among people who liked it |
| Fit | Does this feel like it's for someone like you? | The target segment says no |
Liking is useful as a tiebreaker, not a pass mark. Plenty of well-liked ads fail on linkage.
What questions go in the screen?
- Here is an ad as it would appear in your feed. Would you stop on it, keep scrolling, or skip it?
- In a few words, what was the ad saying?
- Which brand do you think this ad was for?
- After seeing this, which of these would you do? [Search for it, visit the site, save it, nothing]
- Of the ads you saw, which one would make you most likely to buy?
Show each respondent one creative for questions 1 to 4, then the finalists side by side for question 5. Monadic first, comparison second, so one creative doesn't color the read of the next. It is the same order used in a concept test questionnaire.
What does a six-creative screen look like on a simulated audience?
Hypothetical: a running shoe brand has six short vertical videos for a spring cushioning launch and budget to test two in-platform.
- Audience: US adults 25 to 54 who run at least twice a week, split by people who have bought the brand in the past two years and people who haven't, and by social platform used most.
- Design: each creative fielded monadically to a simulated audience, the questions above, then a forced choice among the top three.
- Readout: a scorecard by creative and segment, coded takeaways, and the share who named the right brand.
Example output layout (illustrative, not a real result)
| Creative | Would stop | Correct takeaway | Named the right brand | Would act | Advance? |
|---|---|---|---|---|---|
| Video 1: athlete close-up | __% | __% | __% | __% | Yes / No |
| Video 2: runner POV | __% | __% | __% | __% | Yes / No |
| Video 3: product spin | __% | __% | __% | __% | Yes / No |
| Video 4: founder story | __% | __% | __% | __% | Yes / No |
| Video 5: user testimonial | __% | __% | __% | __% | Yes / No |
| Video 6: humor skit | __% | __% | __% | __% | Yes / No |
If the humor skit scores highest on "would stop" but lowest on brand naming, it either gets a logo in the first seconds and a re-screen, or it is cut. The two that advance go to a live in-platform split, where the growth team's A/B slots are spent only on contenders.
What should you do with the results?
- Fix before you cut. A creative with a strong hook and weak linkage often needs an earlier brand cue, not a new concept. Re-screen the edit.
- Check the message, not just the ad. If every creative produces a scattered takeaway, the problem is upstream in the message; run a message test first.
- Read by segment. A creative that wins with lapsed buyers and loses with current ones may belong in a different flight. The channel planning guide covers where each audience sees ads.
- Keep a live check. In-platform results are the record once money is spent. Treat the simulated screen as the filter, and check how the creative affects brand measures after the flight.
Questions about testing ad creatives
How many ad creatives should you test at once?
Screen as many as you have, often six to twelve, then send two or three to live tests. Live tests split budget, so fewer contenders get cleaner reads.
Can you test ad creative without spending on ads?
Yes, for the screen. Showing creatives to a matched audience before launch costs no media. The live in-platform test still costs spend, which is why the screen comes first. Screen the page the ad points to as well, using a landing page test before launch.
What is brand linkage in ad testing?
The share of people who correctly name the brand an ad was for. Low linkage means the ad may be building memories for the whole category or for a rival.
How does Neon Blue screen creatives?
Neon Blue shows each creative to a simulated audience matched to your buyers, reads hook, takeaway, brand linkage and next action by segment, and marks every result as simulated so it sits apart from live data. Start a free first study.