use case · analyze
Neo tells you which concepts win, which fatigue, and why.
Neo reads your Meta creatives, images and videos, ties each ad to a concept and judges it within its ad set, link by link, from the hook to the landing page. You know what to keep, what to rework and what to produce next.
4,000 advertisers and agencies already use Creaboost.
Every concept gets a status, and every status names the weak link.
Camille asks Neo what works in Maison Lune’s creatives. Neo reads six weeks of Meta, then the sheets of the creatives that carry the spend and of those launched during the week.
Which concepts are winning, which are fatiguing, and why?
Camille Durand · Wed, Oct 7 · 10:20
Neo · 10:31
- Meta · live · 7-day click
- creative reading · 20 ads
- memory · personas and concepts
- 6 weeks · weekly view
Concept 04 shows the fatigue signature, and the founder is breaking out.
Week of 9/28 to 10/4, against the four weeks before: $23,100 spent, $53 CPA in 7-day click. Neo read the ads that make up 80% of spend and every ad launched during the week, then judged each concept within its ad sets. Catalog is read apart, with the audiences.
- spend
- $23,100
- +0.5%
- CPA · 7-day click
- $53
- 4-week avg. $58
- ads read
- 20
- 14 · 80% of spend, 6 new
- concepts
- 5
- catalog apart, 34%
concepts, by status
- breakout6 → 11%
Founder, “a size 8 in pants”: its share rises every week, and its ad set holds.
Live for three weeks. The ad set that carries it runs at $45 in 7-day click, below the account’s reference, with the best link CTR in the account.
- ad set CPA $45
- link CTR 2.5%
- 3 weeks live
- established9%
Founder, “pajamas, commentary”: the best hook in the account, then viewers drop off.
Steady share, ad set at $50. The hook stops more people than anything else; the hold rate shows the bridge to the body of the video breaks right after.
- hook rate 33% · median 22%
- hold rate 3.0% · median 8.7%
- declining25%
Concept 04, “before/after”: the fatigue signature, two weeks in a row.
Frequency up, reach flat, hook rate and link CTR down, CPM and CPA up, while its spend holds. Neo ruled out the look-alikes: account CPM steady, no recent edit, the pieces shown are in stock.
- frequency 2.1 → 3.4
- hook rate −24%
- CPA +38% vs. its best weeks
- in test4%
Maternity, “nothing squeezes”: too early to call.
The best conversion rate in the account and a low hook rate. It has spent $690 of its $1,620 test budget: it gets judged once that budget is spent.
- $690 of $1,620
- conversion rate 1.6%
- hook rate 16%
- failed7%
Period underwear, “demonstration”: test budget spent, reference never reached.
The sheets show an offer, up to 25% off, with no angle. The learning to keep is about the angle, the first thing to change.
- $1,620 spent
- $95 CPA
- offer only
A creative is never judged on its own: hook rate and hold rate per hook, CPA at the ad set level.
missing inputsNo target CPA in Economics, so the reference is the median of the last three months, $55 in 7-day click. Catalog has no single creative to read: it is judged with the audiences.
proposed decisions
- 01
Three new hooks on concept 04’s body.
rule A fatiguing winner keeps its body and changes its opening: Meta counts a new hook as a new ad. If the concept stays worn out, a new concept follows, starting with the angle.
check Oct 20 · hook rate > 26% - 02
Keep the hook of “pajamas, commentary” and rework seconds 3 to 8.
rule A strong hook rate with a low hold rate: the bridge to the body breaks. Change that link, and nothing else.
check Oct 20 · hold rate > 6% - 03
Roll “a size 8 in pants” out as three hooks and a static, in a new ad set.
rule A breakout grows through new hooks and new formats. New ads go into a new ad set; a winning ad set is left alone.
check Oct 27 · ad set CPA ≤ $53 - 04
Stop “demonstration” and record its learning in memory.
rule A concept that spent its test budget without reaching the reference stops, and you keep what it taught you: here, the angle to change.
stop Oct 8 · learning recorded
Saved in Docs · Maison Lune · 10/07/2026
- Export to Google Slides
- Write the 3 new hooks
how Neo gets there
Neo reads the numbers, then the creatives themselves.
- data
Meta live, and the creatives read one by one.
Neo reads Meta results live, then the sheet of each creative, image or video: transcript, hook in the first 3 seconds, angle, format, persona, offer and audio. It reads the ads that make up about 80% of spend, plus every ad launched in the period, and leaves the long tail aside.
- method · meta_pilot
One status per concept: breakout, established, declining, in test or failed.
Fatigue is read over 4 to 6 weeks, and only when its signature shows up two weeks in a row: frequency up, reach flat, hook rate and link CTR down, CPM and CPA up at steady spend. A creative is never judged on its own: hook rate and hold rate per hook, CPA at the ad set level.
- memory · Maison Lune
The concepts in memory give the ads their names.
Neo ties each ad to a concept from Personas & Concepts: persona, angle, offer. An ad that fits none becomes a new concept, flagged for the memory. Your naming conventions and tags are the starting point.
A status for every concept, then the creatives to produce.
The analysis lands in the chat. It becomes a doc, a board that tracks fatigue, or the hooks for the next creatives.
Explore Analyze →docs
The concept analysis, dated.
Exported to Google Slides in the brand’s look, ready for the creative review.
boards
The “Concept fatigue” board.
Frequency, reach, hook rate, link CTR and CPA for each concept, over six weeks.
tasks · monday review
Fatigue, reread every Monday.
The Meta review goes through each concept and flags those whose signature shows up two weeks in a row.
chat
The new hooks, written on request.
Five to eight per concept, on three layers: the visual, the on-screen text, the first line.
go further
More use cases, and the product pages behind them.
questions
What people ask us about creative analysis.
Does Neo really watch the videos?
On Meta, yes: an AI model watches and listens to each creative, image or video, and writes a sheet (transcript, hook, angle, format, persona, offer, audio), computed once and stored. The reading does not cover Google Ads or TikTok Ads, and catalog ads have no single creative to read.
When does Neo call it fatigue?
When the signature shows up two weeks in a row, over a 4- to 6-week window. One week on its own proves nothing. Before concluding, Neo rules out the look-alikes: CPM rising across the whole account, a landing page, a stock-out, a recent edit.
Read: creative fatigue on Meta →Do I need a naming convention?
It helps, but you don’t need one. Neo groups ads by naming convention, by tag groups or with its own tags, and shows you the ones it added.
What does reading creatives cost?
One AI credit per new creative read. A creative already read costs nothing more, even when it runs in several ads.
How is this different from Analyze?
Analyze organizes your creatives into reports your team browses and shares. Neo starts from the same data and answers your question: what wins, what fatigues, why, and what to produce next.
Explore Analyze →See your concepts ranked, on your own Meta account.
Book a demo: we connect your Meta account, and Neo reads the creatives that carry your spend in front of you.