.@DraftKings round 2 β same campaign, different leak: the background texture is fighting your own message. Clarity 69, βbusy.β
A quieter ground lifts every other score on the canvas.
The fix takes an afternoon. Seeing it takes a minute: https://t.co/zuZcytOykP
#AdCreative
Six furniture brands are deep in Labor Day sale creative, and almost none of it is a month old.
We pulled 794 live Meta ads and scored 72 of them. Backcountry and Crate & Barrel have replaced 100% of their ads inside a month, Raymour & Flanigan 97%. Nobody is sitting on a proven winner.
The scores barely move for it: 84 to 87 across every account.
@estemzzei you can see the difference from outside, in the ad library: Backcountry launched 97 ads in the last 30 days and none is older than 29 - Dutch Bros launched 90 and still keeps one from 76 days ago. same testing volume, only one of them is keeping what it learned.
the libraries of brands past that line agree with you from the outside: Dutch Bros has 1,282 ads live and still keeps one running 76 days, Backcountry cycles its whole library inside 30. neither looks like 200 tests - it looks like volume on the few things that already convert cold.
good instinct breaking it down past the creative. the next step that changes the read: open the advertiser's whole library, not the one ad - how many are live right now and how long the oldest has run tells you whether this one is a test they're about to swap or the one they keep paying for. one ad is a snapshot, the library is the strategy.
the 60-day check is the one that separates a test from momentum. two pulls from this morning as calibration: Backcountry - 130 ads in the library, all 97 I sampled still live, oldest 29 days, 97 of 97 launched in the last 30. Dutch Bros - 1,282 in the library, median age 34 days, oldest 76. same check, opposite answer: one is still searching, one has found things it keeps paying for.
the ad library shows this from the outside too. pulled Dutch Bros this morning: 1,282 ads in the library, sampled 200 - median age 34 days, oldest 76, and 90 of the 200 launched inside the last 30. the ones surviving past 60 days are clearly not being judged on last-click, or they would have gone out with the weekly swap.
client brief: make the sale message pop. the model: "END OF SEASON" is never reached. the reason: someone set it on the loudest dress in the shoot. the dress won. the dress always wins.
We measured the meal-kit category's live Meta ads. 63 of 64 creatives carry a structural flaw.
Home Chef 21/22
@HelloFresh 19/19
EveryPlate 15/15
Sunbasket 8/8
An entire category leaking attention at once. Receipts: https://t.co/JTkNNhWkAL
#MealKits
.@BestBuy ships 135 new ads a month β real velocity. But all 4 hero creatives are flagged, including the 101-day workhorse.
Speed times leaky structure is a tax on every impression.
The fixes take one minute each: https://t.co/ykGuPQlycb
#Retail
where does the first glance land on this Crocs concept static? ran it through an attention model (predicted eye-tracking, heatmap attached). the CROCS wordmark takes 67% of all attention and is seen first. the small logo on the giant shoe is seen second, about 6x what its size would earn. the shoes on her feet are never reached. clean canvas, strong anchor - and the product is the one thing the eye skips. if this were yours, would you calm the wordmark or move the product up into the gaze path? context: this is an AI-made concept (the frame says "designed with ChatGPT"), not a Crocs campaign - it is from a sample set I use for testing.
what I'm working on: an attention heatmap for ad creatives. drop a static in, get a predicted first-glance map, three scores, and a plain-English read on what the eye skips. sample below is a Nescafe static from our demo set: the headline pulls 24% of attention at about 5x its size, the jar is seen second, and the subhead under the headline holds 21% but the gaze path never actually reaches it. building it for designers and media buyers who want a second opinion before a creative ships. question for the builders here: if a tool told you this about your ad, what is the one thing you would want it to say next?
We measured 34 kickoff creatives across the fantasy category. 76% are deliberately dense β walls of text, offers, logos, legal.
It's a convention, not a mistake. But when everyone shouts, the calm ad owns the feed.
Who breaks ranks first? π
#NFLKickoff#FantasyFootball
@BeatriceReplo Very clean! Hero sections are extremely important. You barely have 1-2 seconds to catch a user's eye and make them want to scroll. I ran this on our attention heatmap tool in https://t.co/8SuWfhZn66 to understand its insights, attached a part of the results!
@Abedny477177896 Resonating with the users is ultimately what makes all the differnece π€ I ran one of the creatives on our heatmap insights tool in https://t.co/8SuWfhZn66. The beach scene with the people and product cans holds the most attention overall, with 59% of the visual interest.
@DilanOktay80 These caption the vibe so well. I ran the first image on our heatmap insights tool in https://t.co/8SuWfhZn66. The three product images below the headline capture the most attention overall, with 29% of the initial gaze.
@OlaitanRodiah These are clean! An important part of creative ads is checking out competitors' ads to analyse how they approach the situation. I ran a pull for 'face and scalp sunscreen' on our Meta Ad Library tool in https://t.co/8SuWfhZn66 and got some great insights along with the ads!
same. the longest runner is the one signal the budget already voted on. the wrinkle we keep seeing when the long runners go through the attention model: proven and clear are different axes. Dutch Bros' oldest live ad is 60 days old and the account median still sits at 78, while a 111-day beauty static we ran came in at 94. if there is a brand you buy against, name it and I'll pull its longest runners.
the same one who ran your after-photo static last week and sent the link when you asked for it :) I scan ads through the attention model at jarville - you posted a new set and I could not help myself. the card is yours if you want it, and if you'd rather client work stay off a public scoreboard just say so
three statics, one client, ranked by where the first glance lands: #1 (the loop) 88, #2 86, #3 78. same pattern across the set - the illustration and the headline share the eye, and in #3 the CTA gets 0% so anything above it wins by default. winner's full card below, the other two as heatmaps
easiest way is to run one yourself - drop the design on the page and the first read is free, signing in gives you 50 credits so you can batch a set and compare them side by side: https://t.co/0PQJstQst7 glad it was useful. tag me when the next hero ships and I'll run that one too.