Creative Strategy

Know which ads are winning, which are dying, and what to brief next

Drew grades every Meta creative against your own account, reads what each one shows and claims, groups your ads into the concepts they really are, and watches what your competitors launch — then turns the verdict into an action or a brief.

Meta Ads today · graded against your own account on Meta-reported signals, labelled as such · Last updated: September 2026

Get started for free Login

Creative Intelligence

See which ad ideas win — and what to make next

Roughly one ad in twenty becomes a winner, and winners fatigue within weeks, so the loop a creative strategist runs by hand never stops: grade, diagnose, read, compare, brief. The Creative Intelligence section in Datadrew runs it every week — no naming conventions, no spreadsheet archaeology.

CREATIVE LEADERBOARD

Every creative graded, with the evidence behind the call

One row per creative, not per ad. A creative running under six ad sets is counted once, so its ROAS and hook rate are measured on all of its delivery. Each one is graded Scaling, Healthy, Fatiguing, Dead or Testing against its own baseline, and every grade maps to one verb.

Spend, ROAS, link CTR, hook rate, hold rate, frequency and purchases per creative — Meta-reported, and labelled that way
Five states, one verb each: SCALE, ITERATE, KILL, WAIT — and Healthy, which needs nothing from you
Winner badge when spend is at least 10× your account-median creative spend, above the spend floor, with ROAS at or above your account average
The evidence inline — “link CTR −34% vs peak, frequency 3.1” — never a label without the numbers behind it

Creative leaderboard Last 30 days · Meta-reported

Creative	Status	Action	Spend	ROAS	Hook
Summer UGC Winner	Scaling	SCALE	$4,820	5.4x	42%
Before/After Static	Healthy	—no action	$3,190	4.1x	—
Founder Story Reel	Fatiguing	ITERATE	$2,740	2.3x	24%
Brand Montage	Dead	KILL	$1,980	0.7x	14%
Problem/Solution v3	Testing	WAIT	$310	—	—

Founder Story Reel · Link CTR −34% vs 7-day peak · Freq 3.1 · Still 2.3× ROAS · Refresh the hook, don’t kill it

WEEKLY DIAGNOSIS

This week’s actions, ranked by spend at risk

Every week Drew lists the creatives that need a decision, worst first, with the numbers that triggered each call. Testing creatives are held out until they clear the sufficiency gate, and when nothing needs action it says so — no filler recommendations.

KILL, ITERATE and SCALE cards ranked by the spend riding on each verdict
Testing velocity: your launches per week against the peer median and top quartile of Datadrew’s active Meta ad accounts in your spend tier
Creative mix: how many of your creatives sit in Scaling, Healthy, Fatiguing, Dead and Testing right now
“Nothing needs action this week” is a real answer, not an empty state

This week’s actions Ranked by spend at risk

KILL

Brand MontageCPA +58% vs its own baseline · Freq 4.2 · 0.7× ROAS

$1,980at risk

ITERATE

Founder Story ReelLink CTR −34% vs peak · Freq 3.1 · still 2.3× ROAS

$2,740at risk

SCALE

Summer UGC5.4× ROAS at freq 1.8 · CTR at its peak · headroom

$4,820spend

Testing velocity Compared with 212 accounts in your spend tier

You

3 / week

Peer median

5 / week

Top quartile

9 / week

You launch below your peer median. At roughly a 5% winner rate, volume is the most reliable lever you have.

AI-READ CREATIVES

Drew reads every creative — what it shows, what it claims, who it speaks to

No naming conventions required. Drew looks at the image or the video frames and the copy of every creative that has spent enough to judge, tags it on 20 attributes — angle, hook type, visual device, tone, offer, production style, proof devices and more — and writes a plain-language read-out of the ad.

Group performance by angle, hook, visual device, tone, offer, production style or CTA: do social-proof hooks beat offer-led hooks on ROAS?
Cross two attributes — angle × hook is often the most useful creative question there is
Per creative: the claims it makes, the objections it answers, who it appears to speak to, what’s on screen and every word of on-image text
Tag coverage stated on every view, so a rollup never pretends to cover spend it hasn’t read

Group by: Angle 31 of 34 creatives tagged · 96% of spend

Angle	Creatives	Spend share	ROAS	Hook
Problem / solution	9	38%	4.6x	39%
Social proof	7	24%	3.9x	35%
Offer / discount	8	21%	2.4x	22%
Lifestyle aspiration	4	11%	1.8x	19%

Summer UGC AI-read

Claims it makes: “Visible results in 14 days” · “Sold out 3× this year”
Objections it answers: Works on sensitive skin · Free shipping over $60
Speaks to (inferred): Women 30–45 with reactive skin
Text on screen: “Day 1 | Day 14 | still can’t believe it”

CONCEPTS

48 ads. 6 concepts. Now you know how many ideas you’re really testing.

Variants of the same idea count once. Drew groups creatives whose message and visuals are near-identical into named concepts, so you can see where spend is concentrated, which idea is carrying the account, and whether a “new” launch is actually a new argument.

Concepts named from the ads themselves — angle and hook: “Product Feature — Bold Claim”, “Offer Discount — Offer Led”
Spend share, purchase value and ROAS per concept — Meta-reported
“Doesn’t match an existing concept” flagged on its own — often a genuinely new angle
Coverage of your Meta spend stated on the page, never assumed

Concepts Last 30 days · 94% of Meta spend covered

6 distinct concepts behind 48 ads

Problem/Solution — Problem Callout

34% · 4.8x

Social Proof — Testimonial Quote

27% · 4.1x

Offer Discount — Offer Led

19% · 2.2x

Education/How-to — Demonstration

9% · 3.0x

Not yet grouped: 11% of spend — 2 creatives that match nothing else you run. Often a genuinely new angle.

RIVALS

See what competitors are testing — and the angles you have no answer to

Track up to five competitor brands straight from Meta’s public Ad Library, each in the market you choose. Drew collects their creatives, deduplicates them so one creative is one row instead of the dozens of ad IDs Meta runs it under, compares them week over week, and keeps the media so it still loads a month later.

What’s new and what stopped in the last 7 days, per brand
Long-running badge: live 30+ days — the operators’ own proxy for “this is working for them”
Format mix, calls to action and the creators they whitelist — so you can see an angle, offer or format they have moved into that you have no active answer to
A Competitor Watch report you can schedule for Monday mornings — honest about what Meta doesn’t publish: no spend, no results, no estimates

Rivals Meta Ad Library · United States

Gymshark Alo Yoga Vuori + 2 slots left

Last 7 days · 6 new · 2 stopped · 61 active creatives

NewVideo · Reels

“Your gym kit called. It wants a promotion.”

Running as 14 ads · CTA: Shop now

Long-running · 47 daysStatic

“Free shipping over $75 — this week only.”

Running as 6 ads · Creator ad

Public Ad Library data. Meta publishes no spend or results for competitor ads — none is shown or estimated.

The grading contract

Five states, one clear call each — graded against your own account, not a universal cutoff

The same contract runs in the dashboard, in Drew’s answers and in the weekly diagnosis, so a creative is never “fatiguing” on one screen and “healthy” on another.

ScalingSCALE

A fresh winner with headroom

Spend at least 10× your account-median creative spend and above the 500 spend floor, ROAS at or above your account average, frequency under 2.5 and link CTR not declining. Push budget while the signals hold.

Healthy—

At or above its own baseline

No verb, because no action is needed — keep watching frequency. When signals conflict, Drew says “mixed signals — watching” rather than forcing a verdict.

FatiguingITERATE

Proven concept, tired execution

Fatigue confirmed but ROAS still holds. Refresh the hook and the first frame — don’t kill it. This is the verdict that turns into an iteration brief.

DeadKILL

Weak concept — free the budget

Fatiguing and ROAS below the greater of 1.0 and 0.75× your account average. Pause it and move the budget to your scaling ads — don’t re-cut it.

TestingWAIT

Below the sufficiency gate

Under 500 in your ad account’s currency, or fewer than 5 delivery days. Not enough signal to rank, crown or cut — let it run.

Fatigue is a two-condition test

Link CTR down 30% or more from its own 7-day peak and frequency at 2.5 or above — or the recent 7-day cost per purchase at 1.5× the creative’s own 30-day baseline while spend is sustained. A CTR dip on its own is not fatigue, and Drew never calls it on one leg. Nor is a creative ever killed on CTR and frequency alone: Dead needs its ROAS to have collapsed too, so a narrow-audience ad that still converts stays an ITERATE, not a KILL.

Meta-reported, and labelled that way

ROAS, purchases and purchase value are the platform’s own attribution — a leading indicator of creative strength, not a revenue claim. Hook rate is 3-second views ÷ impressions; hold rate is 15-second views ÷ 3-second views. The formulas are frozen, so the number in Datadrew reconciles with the one Drew quotes. Every verdict is relative to your own account’s baseline, so the comparison holds even when Meta’s absolute numbers run hot — and Drew reads your Shopify orders and blended ad efficiency alongside them before it proposes a card.

From verdict to action

A verdict is only useful if something happens next

Drew doesn’t stop at the dashboard. Each call becomes a card you can approve, a brief you can hand over, or an answer where you already work.

Approve it from the board

Drew proposes the pause or the budget change as a recommendation card with the evidence attached. Approve it and Drew executes within your guardrails. Execution is rolling out now.

Brief the next one

Ask Drew for the iteration brief: keep the concept that’s winning, change the hook, and cite the tags, claims and objections that made the parent work. Written for your editor — Drew doesn’t generate images or video.

Where it reaches you

The Creative Intelligence section in Datadrew, Drew in Slack, a Competitor Watch report you can schedule, and every creative tool through Claude or ChatGPT via MCP.

By the Numbers

Creative Strategy by the numbers

0K+

Meta creatives tracked
across Datadrew brands

Attributes read on
every tagged creative

Competitor brands tracked
per shop, in the markets you choose

Drew AI

Ask Drew AI about your creatives

Everything on this page is available conversationally, on the same grading contract. Ask in Datadrew, in Slack, or from Claude and ChatGPT through MCP — and get the verdict, the evidence and the next step in one answer.

Which creatives should I scale, and which should I kill?
Do social-proof hooks beat offer-led hooks on ROAS?
What did my competitors launch this week?
My problem/solution ads win — write the brief for the next one

Learn about Drew AI

Drew AI Ads Agent

Which creatives should I scale, and which should I kill?

Graded on the last 30 days, Meta-reported:

SCALE — Summer UGC: 5.4× ROAS at frequency 1.8, $4,820 spend, link CTR still at its peak.
ITERATE — Founder Story Reel: 2.3× ROAS, but link CTR is down 34% from its 7-day peak at frequency 3.1. Proven concept, tired hook.
KILL — Brand Montage: CPA +58% vs its own baseline, 0.7× ROAS, frequency 4.2.

Proposed: pause Brand Montage and move its $66/day to Summer UGC — two cards on your board, ready to approve. Want the iteration brief for the Founder Story hook?

Know what to brief next

Connect Meta and Shopify. Your creatives are graded and your rivals collected within a day — free to get started.

Start for free Book a demo

Frequently asked questions

What ad platforms does Creative Strategy support?

The graded creative library, AI-read tags, concepts and Rivals cover Meta Ads (Facebook and Instagram) today. Connect your Meta ad account and the first sync of creative data completes within a day, with no manual uploads or naming conventions. Drew works from the Meta history already in your account, so verdicts appear as soon as that first sync completes for every creative that clears the sufficiency gate; thinner creatives stay in Testing rather than being judged on noise. Google Ads responsive search ad and Performance Max asset performance is available through Drew conversationally.

How does Drew know what a creative shows and says?

Drew reads the image, or up to four frames of the video, together with the headline and primary text, and tags the creative on 20 attributes: angle, hook type, visual hook device, emotional tone, offer, production style, proof devices, claim risk and more. It also writes a plain-language read-out of the ad: the claims it makes, the objections it answers, who it appears to speak to, and every word of on-image text. Tags are validated against a human-graded sample and labelled AI-read in the product, so you can always tell which numbers come from Meta and which from the model; coverage is stated on every rollup, and a creative is read once it has spent enough for its performance to mean something.

What is hook rate and how is it measured?

Hook rate measures how many people stopped scrolling to watch your ad: 3-second video views divided by impressions. Hold rate measures how many of those viewers kept watching: 15-second views divided by 3-second views. Together they show whether a creative earns attention and then holds it. The formulas are frozen, so the hook rate on the dashboard, in Drew’s answers and in the weekly diagnosis is the same number.

How does creative fatigue detection work?

Fatigue is a two-condition test, never a single signal. A creative is fatiguing when its link CTR is down 30% or more from its own 7-day peak and its frequency is 2.5 or above, or when its recent 7-day cost per purchase is 1.5 times its own 30-day baseline while spend is sustained. A fatiguing creative whose ROAS still holds is graded Fatiguing and mapped to ITERATE — a proven concept with a tired execution. One whose ROAS has also collapsed below the greater of 1.0 and 0.75 times your account average is graded Dead and mapped to KILL. Every verdict ships with the numbers that triggered it.

Are these numbers Meta-reported or true ROAS?

Meta-reported, and labelled that way everywhere. ROAS, purchases and purchase value are the platform’s own attribution — a leading indicator of creative strength rather than deduplicated revenue. Datadrew does not claim causal credit for every sale. For the business view, Drew reads the Shopify orders and blended ad efficiency behind the ads alongside the creative grades.

What does Rivals show, and where does the data come from?

Rivals is built on public data from Meta’s Ad Library. You track up to five competitor brands per shop, each in the market you choose, because global brands run a separate Page in each country. Drew deduplicates Meta’s many ad IDs down to one row per creative, compares complete snapshots week over week, and keeps a copy of the media so it still loads later. You see what launched and stopped in the last 7 days, which creatives have run 30 days or more, their format, call-to-action and creator mix, and which ads are catalogue templates. Meta publishes no spend or results for competitor ads in any market, so none is shown or estimated. A Competitor Watch report can be scheduled as an automation, Monday mornings by default.

Can Drew act on a creative verdict?

Does Drew write the creative brief?

Drew writes briefs conversationally. Ask for the iteration brief on a winning or fatiguing creative and Drew keeps the concept, changes the hook, and cites the tags, claims, objections and numbers that made the parent work — with hook-body-CTA structure, UGC script beats and one-variable A/B tests. Drew does not generate images or video; the brief is written for your editor or creator to produce.

Is Creative Strategy available on the free plan?

Go deeper