Dara Denney's method for explaining why an ad won — read the campaign then the top spender, tag format, creator, imagery/setting and messaging, ask which awareness stage the message reaches (scalable vs low-hanging-fruit messaging), add breakdowns and comparisons — plus the recurring Motion reports she runs.

Analysing a creative test beyond “variant B won”

TL;DR

A test readout that only says “B won: 1.2 ROAS, 1.64% CTR, 21% hook, 7.8% hold → make more B” is not an analysis — the why lives in the creative, not the numbers.¹²
Tag every tested ad with qualitative data: format, creator, imagery (production quality, setting) and messaging.¹³
Ask which awareness stage each message reaches. In Dara’s test the fastest early converter (low-hanging fruit) was not the long-run winner; the message that reached problem-aware/unaware people scaled longest.⁴
Add breakdowns (age, gender, placement) and comparisons with your own top performers and competitors’ ads.⁴⁵
Don’t dwell on why every loser lost — look for trends in messaging, visuals and formats, then double down and pressure-test them.⁶
Save the deep version for big swings and “unicorn” ads that are scaling, not every test.⁷

Details

Reading a live campaign (Dara Denney, 2025). Sort by spend, look at the campaign as a whole, then zoom in on the ad taking the bulk of spend. In her ASC example one ad took ~75% of spend. Its outbound CTR was only average for the account. Its hook and hold rates were almost double the account averages (~30% hook, 7–8% hold), and that was the real signal.⁸⁹¹⁰ The winner was a pure educational ad that simply explained the product.⁶

Qualitative tags.

Format: e.g. a UGC creator testimonial. Is it similar to other winners, or an outlier?¹
Creator: this creator kept winning across tests, so the brand rebooked her.³
Imagery: production quality and setting. The winning creator filmed in her car. For several brands, shooting away from home has had a big impact on performance.³
Messaging: in the three-way test, A was value-based, numbers-heavy and compared against competitors. B, the winner, used negative marketing, the trigger word “hate” and a question addressed to the viewer. C was fast/easy problem-solution.³

Scalable vs fast messaging. A converted fastest at first because it caught people close to buying. B kept running and won longest because it appealed to the widest, more problem-aware/unaware audience.⁴ See Building Meta creative from personas, angles and awareness levels for awareness levels as a planning dimension.

Breakdowns and comparisons. If a creative over-indexes on 50+, consider more creative for that group. Check Facebook vs Instagram and feed vs stories/reels, and whether formatting explains it. Compare against your own current tests and top performers, and against competitors: what they run, whether they converge on the same lessons, and what they repeat month after month.⁴⁵⁷

Recurring reports (Motion). Dara uses Motion, a sponsor tool, for these reports:

top-performing creatives, weekly and in monthly creative retros with the brand;
one report per individual creative test;
format mix: do we need more images, video, carousels or catalog ads?
a second pass within each format (statics: testimonial, us-vs-them, before/after, celebrity/creator);
agency partners compared (who gets more scope or retainer, who to replace);
creators and celebrities compared (who to rebook, who to find more like).¹¹¹²

You can build these in Ads Manager. She finds Motion easier, and easier to share with teams intimidated by Ads Manager.¹²

Disagreements / caveats

One account, one test, from a sponsored video (Motion). “Hate” and the car setting are observations from her accounts, not rules.
Analysing ad-by-ad sits uneasily with the breakdown effect (The breakdown effect — judge Meta results at the ad-set level, not per ad). Dara uses it to learn patterns for the next brief, not to move budget between ads.

Analysing a Meta creative test beyond "variant B won"

Analysing a creative test beyond “variant B won”

TL;DR

Details

Disagreements / caveats

See also

Analysing a Meta creative test beyond "variant B won"

Analysing a creative test beyond “variant B won”

TL;DR

Details

Disagreements / caveats

See also

Footnotes