Why some ads resist fatigue, and most of an account does not

A durable ad poses a question instead of spending a punchline, rests on a feeling with no expiry date, and is genuinely different from the other seven ads in your account. Here is how each of those is built, the test that catches a one-week ad, and what we cannot prove.

Testing and volume/By Badal Kariwal/01 Sep 2026/10 min read

What is in here

Why do some ads resist creative fatigue?
A question ages differently from a punchline
The four properties that survive a second viewing
What fatigues is usually the account, not the ad
How do you tell fatigue from a bad week?
What we cannot prove about any of this

The short answer

Watch your own ad twice, back to back. If nothing new arrived on the second pass, you have a one-week ad and no budget converts it into a three-month one. A durable ad poses a question rather than spending a punchline, rests on a feeling that does not expire, and uses sound that adds rather than carries. It does not depend on surprise, which is spent by definition. And it differs from everything else in the account, which is why whole accounts fatigue at once even when each ad looks different.

What you get out of this

The four properties that survive a second viewing, each with its test
Why a set of visibly different ads wears out on the same day
The six perceptual axes that replaced our old sameness check
A scored diagnostic to run on a cut you already have
The honest limit: no outcome data, so this is a production argument

Four numbers behind this piece

Where these came from

8finished ads retired in a single verdict: different cuts, identical argument, world and point of viewSutra Haus verdict log

20ads that turned out on inspection to be one ad, twenty timesSutra Haus verdict log

6perceptual axes a set is now measured against. No two pieces may match on more than twoSutra Haus

~50%of creatives are switched off before day 28Motion, Creative Benchmarks 2026

The first three are ours, counted off delivered work in our own verdict log. The fourth is Motion's, from 550,000+ ads across 6,000+ advertisers, and it is linked in the sources at the bottom.

01Why do some ads resist creative fatigue?

Because they were built to survive a second viewing, and most ads are not. Most ads are punchlines: the first watch spends the entire idea and every watch after it is a repeat. A durable ad poses a question instead. It rests on a feeling with no expiry date, it works with the sound off, and it does not depend on you being surprised.

Then there is a fifth reason, and it is not about the ad at all. An account fatigues as a unit. If eight of your ads argue the same thing in the same world at the same energy, a person who saw one has seen all eight, and the set wears out on the schedule of whichever ran first. We know because we built that set and had to retire it.

Five things people believe about fatigue

Flip them

None of these are invented. Four were said to us by somebody paying for creative, and the second one describes a mistake we made ourselves before anyone said it to us.

02A question ages differently from a punchline

A punchline has one job and it does it once. The moment the viewer holds the answer, the ad has nothing to offer them, and the second impression is a cost with no return. This is why a genuinely funny ad can die faster than a plain one: the better the joke, the more completely it is spent.

A question behaves the other way. Every film we ship has to pose one thing a stranger wants answered, and answering material is quarantined until the question has been made to matter. The useful side effect nobody talks about is durability. If the second viewing still has somewhere to go, the ad has not finished paying you.

The same fifteen seconds, spent and unspent

Second by second

Read it as a list

At	Channel	What happens
0.0s	Punchline cut	Setup
1.2s	Punchline cut	The joke lands
2.6s	Punchline cut	Nothing left to find
0.0s	Question cut	Question posed
1.2s	Question cut	Sharpened, twice
6.4s	Question cut	Answered, but only partly
11.0s	Question cut	The detail you missed the first time

Top lane: the idea is fully delivered by 2.6 seconds and the remaining twelve are a recap. Bottom lane: the same assets, with the answer held back and one detail placed late enough that a first viewing misses it. This is a diagram of the structure, not a measurement of two real ads.

The second viewing is the one that decides

We test a hook by showing it cold, once, to a stranger, because a hook is spent after one viewing. Durability is the opposite test, and it is the one at the top of this page. Almost nobody runs it on their own work, which is why an ad's shelf life gets discovered in the reporting instead of decided in the edit.

A hook is judged on the first viewing. Durability is judged on the second, and almost nobody runs that test.
the line we added to our own review sheet

Four cuts that hold their answer back

Ours, made this way

Ephoria - campaign ad

Hotel client - campaign reel

Beauty device - campaign ad

Cafe - studio demo

Every one of these is built on a question that resolves late, and every one still works muted. Watch any of them twice and notice what arrives on the second pass. That is the property this whole piece is about.

03The four properties that survive a second viewing

None of these are craft properties. All four are decided on paper and all four are cheap to check on an ad already running. The fifth belongs to the account rather than the ad, and it is the one people skip. Run them in order: the first failure is usually the whole explanation.

Every one of the four is settled before a frame exists

Where durability is decided

00Story menuquestion or punchline, chosen on paperProperties 1 and 4

01Direction lockthe feeling named in one word, and the sound mode committedProperties 2 and 3

02Mockupwatched muted at feed size, before the full build existsProperty 3, again

03Set reviewthe new ad placed beside the live ones, scored on six axesProperty 5

Nothing on this chain is a rendering decision. By the time footage exists, an ad's shelf life has already been set, which is why fixing fatigue in the edit almost never works.

Run this on an ad you already have live

Tick as you go - it remembers

A question, not a punchline. Watch it twice. On the second pass, is there anything left to find? If the idea was fully spent the first time, this ad has one good week in it.A feeling with no expiry date. Name the feeling in one word. If naming it needs a trend, a sound, or a joke about a format, the ad expires on somebody else's calendar.Sound that adds and never carries. Mute it. If the ad stops making sense, most of the audience never received the ad in the first place, and the ones who did will not receive it twice.No dependence on surprise. Surprise is spent on the first impression by definition. A beat that works only because you did not see it coming has a shelf life of one view.A real place in the set. Put it beside the other four ads running in the same account. If any two make the same argument in the same world, you have fewer creatives than your dashboard says.

Five checks, about six minutes, on a phone with the sound on and then off. A box you cannot honestly tick is the note. You do not need a second opinion to act on it.

The sound check fails in a way that looks like success. A good track carries a weak cut through an internal review, where everyone watches on a laptop with the volume up, and then the ad meets a muted feed and the picture has to do a job nobody built it for. The longer version is in sound is 40% of the ad.

04What fatigues is usually the account, not the ad

We retired eight finished ads in one verdict. They differed on cut rhythm, on transitions, on type and on music. To a viewer they were identical on argument, subject, world, energy, point of view and borrowed genre. We had been measuring editing mechanics, which vary at the click of a menu, instead of perception, which does not.

Twenty earlier ads turned out to share one skeleton: one ad, twenty times. Later, ninety-four static images shared nine camera angles across eighty-five layouts. The same disease in a second organ, three months apart, and on both occasions the person who spotted it was not the person who made it.

Four frames from one account that do not share an axis

Ours, one brand, four worlds

All four are from our own house wellness brand. Product macro on black, street daylight with a person, warm interior with type, and a near-dark still life. Different subject, different world, different energy, different point of view. That is what has to be true before you call four ads four creatives.

We threw away the check that counted cuts and replaced it with one that describes what a person perceives. Six axes, and no two pieces in a set may match on more than two. It is slower, and it catches what the fast check missed entirely. More on setting those axes per brand in one visual language per brand.

Editing mechanics against the six perceptual axes

The check we replaced

Editing mechanics against the six perceptual axes
Dimension	Editing mechanics	Six perceptual axes
What it counts	Cut rhythm, transitions, type treatment, music	Yes. Argument, subject, world, energy, point of view, borrowed genre
How easily a variant changes it	Trivially. A new transition is one click	Yes. Hardly at all. Each axis costs a real decision
What a viewer actually registers	No. Close to none of it	All of it, inside a second
Verdict on our set of eight	No. Eight distinct ads	One ad, eight times
Failure mode	No. Passes a set a stranger cannot tell apart	Refuses a set matching on more than two axes
What it costs to run	Yes. Nothing. It is automatic	Partly. A person has to watch and answer six questions

We do not win the last row. The old check was free and automatic, and the new one costs a person's attention every time a set goes out. We pay it because the free one passed eight ads that a stranger read as one.

The ruleFive variations of one concept is one swing rendered five times. Distinct means different on argument, subject, world, energy, point of view or genre. It does not mean different on transitions, and a dashboard cannot tell the difference between the two.

05How do you tell fatigue from a bad week?

Fatigue is monotonic and slow: the same ad, the same audience, cost climbing across a fortnight. A bad week is abrupt and shared: everything in the account moves together, including ads that launched yesterday. If your newest creative is suffering as much as your oldest, the problem is not the creative. The long version of that call is in telling creative fatigue from a bad week, and the related trap of reading noise as a winner is worth reading beside it.

Will this ad survive its second viewing?

Four questions, scored

The weights are ours and they are opinionated. Treat the band as a conversation opener, not a verdict. The useful output is which question you could not answer well.

06What we cannot prove about any of this

Nothing in our own corpus records what any of these films did in an auction, so this is a production argument about shelf life rather than a measured law about it. Anyone selling you one should be asked for the dataset, and the full version of our own admission sits with the citation audit in where hook rate and hold rate benchmarks actually came from.

The words we use for this, defined

Search the vocabulary

Filter terms

6 terms

Creative fatigueBuying

One ad's cost per result climbing steadily while the rest of the account holds. Slow and specific. It is not a bad week, which arrives on one date and moves everything at once, including yesterday's launches.

The Story-Question lawStory

Every film poses one question a stranger wants answered. Every beat raises it, sharpens it or answers it, and answering material is quarantined until the question has been made to matter.

The six perceptual axesSets

Argument, subject, world, energy, point of view and borrowed genre. No two pieces in a set may match on more than two. It replaced a check that measured editing mechanics and missed sameness completely.

One swingTesting

A genuinely distinct concept. Five renders of one concept is one swing, not five creatives, and the whole group fatigues on the schedule of whichever variant ran first.

The second hookCraft

A deliberate re-hook at the midpoint of any film over about ten seconds, using a different mechanism from the opening. In several films we measured it carries more production value than the opening does.

Surplus multiplierProcess

The build-to-ship ratio, defaulting to two: build ten to send five. It exists so the last creative act is selection rather than correction, which is the only reliable way we know to raise the floor.

House vocabulary, written so you can lift it into your own brief. Where a number belongs to somebody else's research it is attributed in the sources at the bottom of this page.

We also have a specific reason to distrust our own instincts here. A gate we built around change in the first two seconds ranked our accepted films below our rejected ones, and it was withdrawn. Every automated substitute for judgement we have built has failed the same way, which is why the checks above are questions for a person rather than thresholds for a script.

Vote, then see

Last time an ad's cost per result started climbing, what did you change first?

So treat the four properties as a prior. They tell you where to look when an ad dies early, and they are cheap enough to apply before anything renders, which is where the shape of a short ad, second by second gets decided anyway. They do not tell you what will win.

Questions people actually ask

Open what you need

Why do some ads keep performing for months?

Usually because the idea was not spent on the first viewing. An ad built around a question, a feeling that does not date, and a picture that works muted has something left to give on impression number four. The other half of the answer is competitive: it stays fresh partly because nothing else in the account is arguing the same thing at the same energy.

Do durable ads change how often I need to refresh?

They change what you refresh around rather than the cadence itself. An ad built to survive a second viewing has earned a longer run, so it should be the last thing you switch off and the thing the rest of the set gets planned against.

The cadence question itself is answered in full in telling creative fatigue from a bad week, including why a rate beats an interval and why Motion's day-28 line describes what advertisers do rather than what works. What this page adds is which ad in the account you should be most reluctant to lose.

Is creative fatigue the same as audience saturation?

No, and confusing them wastes money in opposite directions. Saturation is a frequency problem and it is fixed by widening the audience or cutting spend. Fatigue is a creative problem and it is fixed by a genuinely different ad. If a fresh creative to the same audience recovers the cost, it was fatigue. If it does not, you were saturating and the new ad just paid the same toll.

Can you design an evergreen ad on purpose?

You can design for durability. You cannot design for a winner. The properties in this piece raise the chance an ad has a second life, and they cost nothing to apply because all four are decided on paper. What none of them do is tell you which idea the auction will reward, which is why we generate a surplus and select rather than defending one concept.

Do longer ads resist fatigue better than short ones?

Length is not the variable. A fifteen-second ad with an unanswered question outlasts a sixty-second one that delivers everything by second four. What longer runtimes buy is room for a second hook at the midpoint, and in several films we measured, the second hook carries more production value than the opening. Room is only useful if you put something in it.

Should I kill an ad the moment cost per result rises?

Not on one bad day. Look for the shape: a single ad drifting upward across ten to fourteen days while its siblings hold is fatigue, and it is worth acting on. A step change across everything on one date is almost never the creative. Check what else changed that day before you throw away the only ad in the account with a reason to last.

An ad wears out because there was nothing left in it the second time, and no media plan fixes that after the fact. A fortnightly refresh schedule guarantees you never find out which of your ads had a second viewing in them.

Where the numbers came from

Motion. Creative Benchmarks 2026: winners are rare - 550,000+ ads, 6,000+ advertisers, roughly $1.3bn of spend. Used for the day-28 figure and the hit rate.
System1. Maximum Star Rating on Test Your Ad Digital - the 1.5% of digital ads reaching a 4-star emotional response, against 10% on TV
Ehrenberg-Bass Institute for Marketing Science. Institute research on distinctive brand assets - the body of work behind distinctiveness being defined against a context, not in isolation

Every figure above links to the place it was published. Numbers marked as ours are measured inside this studio and we say so where they appear. We do not print a statistic we cannot point at.

Badal Kariwal

Runs Sutra Haus, a one-person ad studio that has shipped over a thousand finished creatives - film and stills - for DTC brands and hotels. Writes here about what the work actually taught him, including the parts that failed. The person who reads your brief is the person who builds the work. Send him something to make.

The test, run on your product instead of ours

Send a link. Get one finished ad back.

One finished cut built from your own product, inside three days, free and yours to run whether or not we ever work together. Watch it twice before you decide. The second viewing is the one that tells you what it is worth.

Replies within a day. Ad within three.