The Gauntlet: how ad scripts get killed before they spend

What is the Gauntlet?

A pre-launch kill filter for ad creative scripts. It is the system that kills about 80% of scripts before they ever spend a dollar, so only what survives gets tested live.

It is a methodology I built and use on my own work. It is not software, and it is not something a client buys access to. It is the reason a client on a 10-creative-a-week slab receives ten scripts rather than the forty or fifty I had to write to get there.

What makes it work is the sequencing. Everything in the Gauntlet happens before a script is eligible to spend money, which is the only window where killing something is free.

Why kill scripts before they spend instead of testing more?

The industry reflex is to test more creatives. Volume is the default answer to almost every account problem, and it is the answer I spent years giving.

Here is the arithmetic that changed my mind. A creative test needs four to six weeks before the read means anything. Before that, you are looking at noise and calling it a trend. So a testing queue is not a list of ads. It is a list of month-long slots, and you have a fixed number of them per quarter.

Now put a script into one of those slots that you already suspected was weak. You just spent budget and a month of learning capacity to be told something a careful read-through would have told you for free. Do that with four scripts out of five and you have burned most of a quarter confirming your own doubts.

Testing more creatives makes that worse, because it adds slots at the bottom of the queue while the constraint sits at the top: how good the median script entering the queue is.

The honest version of the same claim: I am not certain the 80% I kill would all have lost. I will never know. There is no counterfactual, because a killed script generates no data of any kind. What I can say is that the queue got shorter, and the scripts left in it were ones I could defend line by line before they cost anyone money.

How do you get from a product to an angle worth testing?

Nothing gets written until four things are done.

The ICP, then the psychological profile of that ICP. Not a demographic. The specific person, what they have already tried, what they are afraid of being wrong about. On high-ticket, the buyer has usually been researching for months before they see your ad.

Market sophistication. Eugene Schwartz's five levels, which are prior art and not my framework. High-ticket "boring" products almost always sit at level 4 or 5, which means the audience has seen every claim in the category already. At that level the claim cannot carry the ad. The angle has to. This single diagnosis kills more scripts than any other stage, because a level-2 claim written into a level-5 market reads as noise to the only people who could afford the product.

The full funnel. Ad, landing page, checkout, upsell or downsell or order bump. Where does the click land and does that page do its job. A script that sends people to a page that cannot receive them is a dead script no matter how good the hook is.

The existing ads, last 7 days and last 30. What angles are already live, and is each ad assigned to the funnel stage it was written for. Half the angle backlog usually gets cut here, because it repeats something already in market.

Only then does the writing start, and it starts with a benefits list rather than an idea.

Every benefit the product has, material and emotional, written out. The material column is easy and mostly useless on its own: dimensions, warranty, delivery time, what it is made of. The emotional column is where the angles come from, and it is the one most briefs skip because it feels soft. For a high-ticket outdoor product, the material benefit is the build spec and the warranty. The emotional benefit is that the buyer stops thinking about it: no more re-researching it every spring, no more explaining the last cheap one to their partner. Those are different ads. They are often different funnel stages.

Angles get picked against a funnel goal, because the funnel goal is what makes an angle valid or invalid, and TOF, MOF and BOF are three different jobs rather than three budgets:

• TOF exists to start a conversation. Landing page, add-to-cart. It is not there to close.
• MOF moves someone from the landing page to initiating checkout.
• BOF converts an initiated checkout into a purchase. On high-ticket this is where most of the sales are, and the shape that wins is short, one benefit, urgent, and fluid.

An angle can be excellent and still be wrong, because it belongs to a different stage than the one the ad was assigned. That is the most common kill I make, and it is the one that looks least like a creative problem from inside the ads manager.

How do you write 10–20 hooks and ship one?

Per angle, ten to twenty hooks. One ships.

You write twenty because hooks eleven through twenty are where you stop writing the obvious version. The first five are almost always the category's default phrasing, which is exactly what a level 4 or 5 audience has already scrolled past.

The selection criteria are narrow:

Shortest. Maximum meaning in minimum time. If two hooks say the same thing, the shorter one ships.

Most fluid. Read it out loud. If your mouth stumbles, it is dead. Fluidity is the words flowing when spoken, and it is a physical test, not an aesthetic one. This is the criterion that kills the most hooks, because a hook can be smart on the page and unsayable in a mouth.

Easiest to understand. Understood on the first pass, at speed, by someone who was not paying attention. Any hook that needs a second sentence to make sense has already lost the person it was written for.

A hook that fails all three does not get rewritten. The angle goes back in the pile and I move on, because a hook that will not go fluid after twenty attempts is usually telling you the angle underneath it is muddy.

Then the body. Several variations, not one. Select the one that reads effortlessly: a few short lines against one long line, no awkward break, nothing that makes the reader's eye stop where the writing did not intend a stop. Same out-loud test.

Then the CTA, matched to the funnel stage. A purchase CTA on a TOF ad is a kill on its own, no matter how good everything in front of it was.

What actually gets a script killed?

Seven checks. A script that fails any of them does not ship.

1. Stage mismatch. The angle belongs to a different funnel stage than the one the ad is assigned to. 2. Sophistication mismatch. The claim is pitched at a level the market has already exhausted. 3. Hook fluidity. It cannot be said out loud cleanly after twenty attempts. 4. Hook comprehension. It needs a second sentence to land. 5. Body rhythm. Awkward breaks, flat meter, a line that makes the eye stop where it shouldn't. 6. CTA mismatch. The ask belongs to a different stage than the ad. 7. Redundancy. The account is already running this angle, so the test would only tell you what the live data already says.

None of these need a dollar of spend to check, which is the whole design. They are all readable off the page.

What does the live test still have to decide?

Everything that matters commercially.

The Gauntlet raises the quality of what enters the queue. It does not tell you what a real audience does with it, and I have been wrong about which of two surviving scripts would win often enough to have stopped guessing out loud.

So: first ads live in week two. Then four to six weeks before the read means anything. The filter buys you a better queue. It does not buy you a shorter test, and any claim that it does would be a claim about certainty that a sample of live ads has not earned.

The honest limit on the number itself: 80% is my own count of my own process, not an audited figure, and it moves by account. In a regulated vertical there is a second criterion sitting in front of all seven checks, because every word is a compliance decision before it is a performance decision, and a script can be killed on that alone.