The Creative Testing System

Pick a product

Same system. Different product.

The steps do not change. The ads, the angles and the numbers do. Choose one and the whole page runs on it: the replay, the ads, the winners and the log.

Creative Operations Aurelia Skin illustrative replay round 1

00Intro

01Read

02Build

03Test

04Decide

05Grow

Printed ad variations laid out on a desk, one circled in green

Paid social · A working system

The Creative Testing System

How one ad becomes seven, seven become twenty one, and the winners keep getting better.

Michael Bennett

Reading the results and the market Round 1 · Monday

Of the budget on one ad0up from 38% in June

Times each person has seen it0past 3 is where ads wear out

People who stop scrolling0in 14 days, the earliest warning sign

Wearing out faster than we are replacing it.Time to build new versions before it stops paying for itself.

MarketTwo competitors switched to customer-style photos this month.

ReviewsBuyers keep saying the same thing: it absorbs fast. Nobody has put that in an ad yet.

Building seven new versions, one change each

CHANGE 01The look

CHANGE 02Proof

CHANGE 03Words and deal

1ad to beat→7new versions→21in the next round

Each ad runs until it has enough sales to judge it fairly Day 1

The target is set per campaign. For one brand it is four sales, for another a hundred. The line on each bar is that number.

AAd to beat0% of target

BLifestyle0% of targetsafety rule

CCustomer photo0% of target

DReview0% of target

EIngredient0% of target

FTimeline0% of target

GOffer0% of target

The rules decide. Two ads clear the bar.

AAd to beattarget reachedto beat

BLifestyle72% of targetsafety ruleKILL

CCustomer phototarget reachedPROMOTE

DReviewtarget reachedPROMOTE

EIngredienttarget reachedKILL

FTimeline58% of targetHOLD

GOffertarget reachedKILL

The winners. Each one seeds its own next round.

WINNER

WINNER

2winners→4new versions live→12coming next

The idea to take with you

Most ad accounts start over every quarter. This one keeps getting better.

One ad to beat. One change at a time. Every winner seeds the next round.

Your system asks for reduced motion, so this starts paused. Press play.

Creative Operations Northlight Gear illustrative replay round 1

00Intro

01Read

02Build

03Test

04Decide

05Grow

Paid social · A working system

The Creative Testing System

How one ad becomes seven, seven become twenty one, and the winners keep getting better.

Michael Bennett

Reading the results and the market Round 1 · Monday

Of the budget on one ad0up from 41% in July

Times each person has seen it0past 3 is where ads wear out

People who stop scrolling0in 14 days, the earliest warning sign

Wearing out faster than we are replacing it.Time to build new versions before it stops paying for itself.

MarketTwo competitors moved to in-hand trail photos this month.

ReviewsBuyers keep saying the same thing: it survived a river. Nobody has put that in an ad yet.

Building seven new versions, one change each

CHANGE 01The look

CHANGE 02Proof

CHANGE 03Words and deal

1ad to beat→7new versions→21in the next round

Each ad runs until it has enough sales to judge it fairly Day 1

The target is set per campaign. For one brand it is four sales, for another a hundred. The line on each bar is that number.

AAd to beat0% of target

BLifestyle0% of targetsafety rule

CCustomer photo0% of target

DReview0% of target

ESpec led0% of target

FTimeline0% of target

GOffer0% of target

The rules decide. Two ads clear the bar.

AAd to beattarget reachedto beat

BLifestyle72% of targetsafety ruleKILL

CCustomer phototarget reachedPROMOTE

DReviewtarget reachedPROMOTE

ESpec ledtarget reachedKILL

FTimeline58% of targetHOLD

GOffertarget reachedKILL

The winners. Each one seeds its own next round.

WINNER

WINNER

2winners→4new versions live→12coming next

The idea to take with you

Most ad accounts start over every quarter. This one keeps getting better.

One ad to beat. One change at a time. Every winner seeds the next round.

Your system asks for reduced motion, so this starts paused. Press play.

Creative Operations Margin illustrative replay round 1

00Intro

01Read

02Build

03Test

04Decide

05Grow

Paid social · A working system

The Creative Testing System

How one ad becomes seven, seven become twenty one, and the winners keep getting better.

Michael Bennett

Reading the results and the market Round 1 · Monday

Of the budget on one ad0up from 37% in July

Times each person has seen it0past 3 is where ads wear out

People who stop scrolling0in 14 days, the earliest warning sign

Wearing out faster than we are replacing it.Time to build new versions before it stops paying for itself.

MarketTwo competitors moved to in-hand street photos this month.

ReviewsReaders keep saying the same thing: they read it twice. Nobody has put that in an ad yet.

Building seven new versions, one change each

CHANGE 01The look

CHANGE 02Proof

CHANGE 03Words and deal

1ad to beat→7new versions→21in the next round

Each ad runs until it has enough sales to judge it fairly Day 1

The target is set per campaign. For one brand it is four sales, for another a hundred. The line on each bar is that number.

AAd to beat0% of target

BLifestyle0% of targetsafety rule

CCustomer photo0% of target

DReview0% of target

ESpec led0% of target

FTimeline0% of target

GOffer0% of target

The rules decide. Two ads clear the bar.

AAd to beattarget reachedto beat

BLifestyle72% of targetsafety ruleKILL

CCustomer phototarget reachedPROMOTE

DReviewtarget reachedPROMOTE

ESpec ledtarget reachedKILL

FTimeline58% of targetHOLD

GOffertarget reachedKILL

The winners. Each one seeds its own next round.

WINNER

WINNER

2winners→4new versions live→12coming next

The idea to take with you

Most ad accounts start over every quarter. This one keeps getting better.

One ad to beat. One change at a time. Every winner seeds the next round.

Your system asks for reduced motion, so this starts paused. Press play.

Creative Operations Fetch & Field illustrative replay round 1

00Intro

01Read

02Build

03Test

04Decide

05Grow

Paid social · A working system

The Creative Testing System

How one ad becomes seven, seven become twenty one, and the winners keep getting better.

Michael Bennett

Reading the results and the market Round 1 · Monday

Of the budget on one ad0up from 39% in July

Times each person has seen it0past 3 is where ads wear out

People who stop scrolling0in 14 days, the earliest warning sign

Wearing out faster than we are replacing it.Time to build new versions before it stops paying for itself.

MarketTwo competitors moved to doorstep delivery photos this month.

ReviewsOwners keep saying the same thing: the bowl is empty in a minute. Nobody has put that in an ad yet.

Building seven new versions, one change each

CHANGE 01The look

CHANGE 02Proof

CHANGE 03Words and deal

1ad to beat→7new versions→21in the next round

Each ad runs until it has enough sales to judge it fairly Day 1

The target is set per campaign. For one brand it is four sales, for another a hundred. The line on each bar is that number.

AAd to beat0% of target

BLifestyle0% of targetsafety rule

CCustomer photo0% of target

DReview0% of target

ESpec led0% of target

FTimeline0% of target

GOffer0% of target

The rules decide. Two ads clear the bar.

AAd to beattarget reachedto beat

BLifestyle72% of targetsafety ruleKILL

CCustomer phototarget reachedPROMOTE

DReviewtarget reachedPROMOTE

ESpec ledtarget reachedKILL

FTimeline58% of targetHOLD

GOffertarget reachedKILL

The winners. Each one seeds its own next round.

WINNER

WINNER

2winners→4new versions live→12coming next

The idea to take with you

Most ad accounts start over every quarter. This one keeps getting better.

One ad to beat. One change at a time. Every winner seeds the next round.

Your system asks for reduced motion, so this starts paused. Press play.

One full round, start to finish. It plays on its own, or step through it with the controls. Everything below is the same process, held still so you can read it.

STAGE 02

Build

The lineup

Every new ad is the ad to beat with exactly one thing changed. Grouped by what changed, so when one wins you know why.

Aurelia, the ad to beat — AThe ad to beat · running now

Kept the same across all eight

The product, the deal, who sees it, where it shows, the budget, and the landing page. If any of those change, it is not a test any more, it is a new campaign.

CHANGE 01The lookHow it is shot

Aurelia ad, customer photo — CCustomer photo

CHANGE 02ProofWhat backs the claim

CHANGE 03Words and dealWhat it argues

Eight is what this budget can carry, not a rule. Split the money too many ways and no single ad gets enough sales to tell you anything. That is not a bigger test, it is the same test with a blurrier answer.

THE LOOP

Weekly

01

Read

Which ads are wearing out, and what the market is moving toward.

→

02

Build

New versions of the best ad. One change each.

→

03

Test

Each ad gets a sales target before a dollar is spent.

→

04

Decide

Rules written in advance say promote, wait or stop. Every winner gets more budget and seeds the next round.

every winner seeds the next round, and 04 feeds 01

STAGE 02b

Copy

Then the words get their own test

Once a picture wins, it stays put and only the words change. This is where the cheap wins are, and the half that usually never gets tested because it does not look like work.

Aurelia ad, the winning picture, held still — CThe winning picture, held still

C1Original wordsI have tried nine serums this year. This is the first one I repurchased before the bottle ran out.

C2LengthNine serums. One repurchase. Here is the one that earned it.

C3HeadlineSame copy. Headline moves from the personal story to 4.8 stars from 10,412 buyers.

C4The buttonSame copy, same headline. Button moves from Learn More to Shop Now.

STAGE 03

Test

Every ad gets a sales target before it goes live

The target is the number of sales an ad needs before its result means anything, and it is different for every campaign: a few sales for one brand, a hundred for another. If an ad cannot realistically get there, it is merged, given more budget, or it waits. That is what stops us from mistaking luck for a win.

BACK THE WINNERS 55% BUILD ON THEM 30% TRY NEW 15%

Back the winners. Every ad that has already earned it, split by how well it is doing. Build on them. New versions of those winners, sharpening what works. Try new. Fresh angles, funded every round whether or not anyone feels like it.

Judge on sales, not on days

A seven day test is a habit, not a rule. An ad is ready to judge when it has enough sales, whenever that is.

judge it when sales reach the target

One change per ad

Change two things at once and you cannot tell which one won, so the next round is built on a guess. Fine to run. Not fine to learn from.

Trying new things is protected

It is the first budget cut when a quarter gets tight, and the reason accounts run out of winners a year later.

STAGE 04

Decide

The rules, agreed before anyone sees a result

Three checks, in this order, run against live results. Writing them down first is the only way they survive a week where someone loves the losing ad.

INPUTEvery ad that is runningChecked against live results all week, not once on Friday.

CHECK 01Broke a safety rule?Shown too often, cost per sale too high, wearing out too fast, or breaks a Meta rule.

yesKILLPaused on the spot, without waiting for Friday. The idea stays open.

CHECK 02Enough sales to judge it?Enough sales to tell a real result from luck.

not yetHOLDKeeps running into the next round. No call made about it yet.

CHECK 03Beats the current winners?Cost per sale against the ads winning right now, not against last quarter.

noKILLPaused. The idea stays open, so next round tries a different version.

OUTCOMEPromoteJoins the winners, gets more budget, and seeds the next round.

Safety rules that stop an ad early

The same people are seeing it too often
Cost per sale climbs past the agreed limit
It is wearing out faster than we can replace it
It breaks a Meta rule or makes a claim we cannot back

The Meta rule is the one that bites

Meta restricts before and after photos and personal health claims for skincare, so ad F is built as a dated timeline instead. An ad that wins and cannot run is not a win.

Every winner raises the bar

A new winner becomes part of what every later ad has to beat, so the account gets harder to beat on purpose.

ROUND 2

Grow

Usually more than one angle wins

This is the part most people get wrong. Two very different ads can both win, and they are not competing, because they are selling to different people. Both get more budget, and each one seeds its own next round.

WINNERround 1 winner

build on it

Won on cost per sale in round 1. Now the question is which person and which setting.

WINNERround 3 winner

build on it

Promoted two rounds later, on the same measure. Reaches a different buyer than the customer-photo angle does.

Aurelia ad, review stack — D1Review stack

+2 new ideas

try new

Two fresh ideas get tested every round out of the try-new budget. An account that only builds on its winners eventually runs out of winners.

Why not just run the best one

One ad carrying the whole budget gets shown to the same people over and over, and that is what wears an ad out. Several winners spread the same spend across more people and buy the account months of life.

Winners compete with each other

The lineup is not a museum. Budget moves between winners every week based on cost per sale, and a winner that fades drops out without a meeting about it.

A new version has to beat the one it came from

Not the original ad. Each round is measured against the version before it, which is how an idea keeps getting better instead of drifting.

OUTPUT

The log

The decision log

Every round ends in a written row. This is what lets you look back a quarter later and see exactly why each call was made.

Round	Ad	What changed	Decided on	Call	What happened next
01	C	Customer-style photo	Cost per sale, target reached	PROMOTE	Joins the winners. Its next versions get built.
02	E	Ingredient led	Cost per sale, target reached	KILL	Idea stays open. Next version leads with one ingredient, not three.
02	D	Review led	Not enough sales yet	HOLD	Keeps running into round 3. No call made yet.
03	D	Review led	Cost per sale, target reached	PROMOTE	Second winner in the lineup. Runs alongside C, not instead of it.
03	C2	Shorter caption	Clicks that turn into sales	PROMOTE	The shorter caption becomes the default on both winners.
01	C	Customer-style photo	Cost per sale, target reached	PROMOTE	Joins the winners. Its next versions get built.
02	E	Spec led	Cost per sale, target reached	KILL	Idea stays open. Next version leads with one spec, not three.
02	D	Review led	Not enough sales yet	HOLD	Keeps running into round 3. No call made yet.
03	D	Review led	Cost per sale, target reached	PROMOTE	Second winner in the lineup. Runs alongside C, not instead of it.
03	C2	Shorter caption	Clicks that turn into sales	PROMOTE	The shorter caption becomes the default on both winners.
01	C	Customer-style photo	Cost per sale, target reached	PROMOTE	Joins the winners. Its next versions get built.
02	E	Spec led	Cost per sale, target reached	KILL	Idea stays open. Next version leads with one claim, not two.
02	D	Review led	Not enough sales yet	HOLD	Keeps running into round 3. No call made yet.
03	D	Review led	Cost per sale, target reached	PROMOTE	Second winner in the lineup. Runs alongside C, not instead of it.
03	C2	Shorter caption	Clicks that turn into sales	PROMOTE	The shorter caption becomes the default on both winners.
01	C	Customer-style photo	Cost per sale, target reached	PROMOTE	Joins the winners. Its next versions get built.
02	E	Spec led	Cost per sale, target reached	KILL	Idea stays open. Next version leads with one ingredient, not six.
02	D	Review led	Not enough sales yet	HOLD	Keeps running into round 3. No call made yet.
03	D	Review led	Cost per sale, target reached	PROMOTE	Second winner in the lineup. Runs alongside C, not instead of it.
03	C2	Shorter caption	Clicks that turn into sales	PROMOTE	The shorter caption becomes the default on both winners.

Read rows two and three together. An ad that has not earned a call does not get one, and an idea that loses is not dead, it comes back as a different version next round.

CADENCE

The week

The weekly routine

A process that only runs when someone remembers to run it is not a system.

MonTueWedThuFriSatSun

MonReadCheck what Meta is reporting and what the market is doing. Decide which ideas to test this week.

TueBuildMake the new versions, then send them for brand and claims sign-off.

WedLaunchOnly what was signed off, and only with a sales target already set.

Thunothing scheduled

FriDecideJudge every ad that reached its target. Promote, wait or stop, and write it down.

Ads run and collect sales toward their target

Safety rules watch live results all week, including Saturday. An ad that breaks one is paused the moment it happens, not on Friday.

Friday’s winners are Monday’s starting point

DELTA

vs usual

What actually changes

	The usual process	Run as a system
Variation	One idea, a few versions, all different in several ways at once	One change per ad, so you know what caused the win
Copy	Written once with the picture, rarely tested on its own	Its own test: caption, headline and button change one at a time
Targets	Decided after launch, if at all	Every ad gets a sales target before it goes live
Judging	A weekly look, often before there are enough sales to mean anything	Judged when the target is reached, whenever that is
Deciding	A discussion, and whoever feels strongest	Rules agreed before anyone sees a result
Carrying forward	One hero ad, rebuilt from instinct when it tires	Every winner gets more budget and seeds its own next round

Next step

See it run against a real account

The parts worth judging are the ones that touch your data. I will look at a live account and show you which ads are wearing out, which ideas are worth testing first, and what a fair test looks like at your sales volume.

Request the audit →

One ad to beat. Seven new versions. One change each.

Same system. Different product.

The Creative Testing System

The Creative Testing System

The Creative Testing System

The Creative Testing System

The lineup

Read

Build

Test

Decide

Then the words get their own test

Every ad gets a sales target before it goes live

Judge on sales, not on days

One change per ad

Trying new things is protected

The rules, agreed before anyone sees a result

Safety rules that stop an ad early

The Meta rule is the one that bites

Every winner raises the bar

Usually more than one angle wins

Why not just run the best one

Winners compete with each other

A new version has to beat the one it came from

The decision log

The weekly routine

What actually changes

See it run against a real account