Hook rate
did the first three seconds earn the view. Below the floor, nothing after it matters; the fix is a new opening, not a new body.
AI PERFORMANCE CREATIVE
Creative judged on CAC and ROAS, not applause.
We run creative like a testing machine: every asset ships with a hypothesis, gets tested in-market, and either scales or dies on the numbers. AI gives us the volume to test relentlessly; the results tell us what to make next.
Every batch
ships with a written hypothesis before anything is produced
7-day click
the attribution window we read considered purchases on, so winners are not mistaken for losers
Weekly
the cadence of new creative entering the account, so fatigue is answered before it shows
Trusted by brands that live and die by creative performance.
Pretty work that never gets measured is a cost, not an asset. And without volume, you can't test enough to find the winners. Performance creative fixes both, a system, not a one-off.
This is the methodology behind our AI creative agency.
Performance creative is advertising creative built and judged to move a business number: hook rate, click-through, cost per acquisition, return on ad spend. It is produced in volume from explicit hypotheses, tested in-market under fixed rules, and then scaled or retired on the results. A performance creative agency runs that loop as a system, week after week, and owns the numbers.
Not brand creative with a pixel attached. The hypothesis comes first.
Not a one-time batch. The system is the product; the assets are its output.
Not media buying. Media decides where spend goes; performance creative decides what earns it, and the two work best in one loop.
01
every batch tests a specific bet
02
AI produces enough to test properly
03
hook rate, CTR, CAC, ROAS on the metric ladder
04
winners get fresh iterations
05
losers logged, never re-tested
06
the account gets smarter every week
The loop
At the center of the loop sits the database: every concept tested, every result logged, feeding the next hypothesis.
Every creative is read on the metric that matches its job, in this order, before any budget decision is made.
did the first three seconds earn the view. Below the floor, nothing after it matters; the fix is a new opening, not a new body.
did the middle keep them and did the ask move them. A high hook with a low click points at the claim or the offer.
the money metric, read only once a creative has cleared the spend floor and the read window.
which assets are earning their budget, not just the campaign average that hides them.
how fast the numbers above degrade as the same people see the ad again, which sets the refresh date.
A creative is not a loser after a hundred dollars. The spend floor and the read window are fixed before launch so opinion cannot move them afterward.
| Metric | Job of the read | What we change |
|---|---|---|
| Hook rate | Did the opening earn the view | The first three seconds: new opening, new visual, new first line |
| Hold rate | Did the middle keep attention | Pacing, proof order, the length of the middle |
| CTR | Did the ask move them | The claim, the offer, the call to action |
| Cost per result | Is the asset paying for itself | Budget: scale, hold or retire once the read window closes |
| Creative-level ROAS | Which assets earn their budget | Spend split between winners, iterations and challengers |
| Frequency | How fast the numbers decay | The refresh date and the next batch entering the account |
Because we make and run the creative, the performance loop is closed, insights from the media feed the next brief. See how we launch and scale it, and where the assets themselves are built in our AI ad creative work.
An account finds a winner, makes variations of it, concentrates spend on it, and for a while everything looks great. Then it fatigues, and because nothing else was ever funded enough to be judged, the whole account drops with it. We have watched this pattern take accounts from a strong return to underwater in a quarter. The fix is structural: no creative carries more than a set share of spend, and there are always funded challengers in the account.
A brief that does not state what it is testing cannot be read. Ours name the bet in one sentence: problem-first openings beat aspirational ones on cold audiences; product in hand by the third second lifts click-through; a price anchor in the hook raises cost per result on warm traffic. Win or lose, the result goes in the database, and no concept is tested twice.
Most creative hit rates sit in the single digits. At that rate, testing two assets a month finds a winner a few times a year; testing ten a week finds several a month. AI changes the cost of volume, which changes how often you find a winner and how much runway you have before the current one fades. That is the whole reason AI belongs in performance creative.
When a creative works, the highest-return move is not a new concept. It is new hooks on the proven body, new openings on the proven script, new formats and new presenters on the proven concept. Exploitation carries the account while fresh concepts search for the next winner. Both run at once; exploitation gets the larger share.
Fragmented tests, dozens of small ad sets each starving, produce noise that looks like failure. We consolidate testing into fixed drops with a fixed spend per creative and a fixed read window. We read considered purchases on a seven-day click window, because a shorter window under-counts a purchase that takes a few days to make and turns real winners into apparent losers. Top-of-funnel creative is judged on hook rate, reach and the lift it gives the ads below it; bottom-of-funnel on direct cost per result. Two yardsticks, applied on purpose.
Several funded winners live at once, none carrying more than its share. New creative entering every week. Cost per result flat or falling as spend rises. And a database of tested concepts that makes every next brief smarter than the last.
How many proven creatives are carrying spend at once, so no single asset holds the account up.
The concentration check: the leading asset stays inside its ceiling while challengers stay funded.
The cadence of fresh assets entering the account, so fatigue is answered before it shows.
Whether cost per result holds flat or falls as spend rises, which is the only proof the loop is working.
The internet is filling up with AI slop. We don't ship it.
Every asset is human-directed, briefed by strategists, refined by creatives, approved before it runs.
On-brand, always, your voice and guidelines are the guardrails the AI works inside.
Accountable to performance, if it doesn't earn its spend, it doesn't stay live.
By spend and cadence: monthly media spend sets the weekly creative volume the account needs, the mix of concepts and exploitations follows the database, and the loop closes fully when WKND also runs the media. Priced before work starts.
Hypotheses, production volume and the read method, delivered into the media team you already have.
We produce the creative and run the tests, so the read and the next brief sit with one team.
Prospecting through retention: creative, media and the database that makes every next batch smarter.
| AI tools (DIY) | Traditional agency | WKND | |
|---|---|---|---|
| Speed to creative | Fast, but you do it | Slow, days to weeks | Fast, hours to days |
| On-brand quality | Inconsistent | High | High, human-led |
| Volume at scale | Limited by you | Expensive | High, cost-efficient |
| Runs the media too | No | Rarely | Yes, we run and optimize it |
| Accountable to ROAS | No | No | Yes |
Three outcomes decide whether the loop is working: spend spread across proven creative, a read the team trusts, and an account that learns faster than it fatigues.
Spend spreads across several proven creatives with fresh iterations behind each one, so no single asset carries the account.
Concentration risk designed out
Fixed spend floors and fixed read windows agreed before launch, so no creative is killed on a hunch or saved on a preference.
Rules set before the numbers arrive
Every tested concept is logged win or lose, so the database, not memory, decides what gets made next week.
Learning kept, never repeated
The working relationship
Spend floor, read window and refresh date are fixed up front, so nobody argues a creative back to life afterward.
The people making the creative are the people reading the account, so nothing gets lost in a handoff.
Every tested concept is logged with its result, so next quarter's briefs start from what is already known.
Book a creative strategy call, we'll audit your current creative and show you what AI-native, performance-led production looks like for your brand.
A performance creative director who writes the hypotheses and reads the account, a creative team producing the volume, and a media lead who runs the tests and feeds the results back. One loop, one set of numbers.
Writes the hypotheses and reads the account: what each batch is testing and what the numbers say about it.
Produces the volume the testing needs, on-brand, at the cadence the account's spend and fatigue curve demand.
Runs the tests in platform and feeds the results back into the next brief. One loop, one set of numbers.
Creative built and judged to drive a business result, hook rate, CTR, CAC, ROAS, rather than to win awards. It's produced, tested, and iterated as a system.
AI gives us the production volume to test far more concepts, so we find winners faster and refresh before fatigue sets in.
The data. Winners get more spend and fresh iterations; losers are logged and dropped. No opinions overriding results.
Both, that closed loop is the point. We optimize in-platform and feed the learnings back into the next creative batch.
Enough to keep a hypothesis-driven pipeline running weekly, so your account never fatigues unaddressed.
Book a creative strategy call, we'll audit your current creative and show you what AI-native, performance-led production looks like for your brand.