The strategist has two jobs

The problem

Most people working on short-form video are good at one of two things. Some are creators. They know how to make someone stop scrolling, they know when a video is boring, and they can feel that the pacing is wrong before they can explain why. Some are buyers. They live in numbers. They can look at a table and tell you which video is making money, which one is leaking, and roughly where.

A creative strategist has to be both. And almost everyone who comes into this job comes in from the creator side.

If you only think like a creator, you can make something people love that makes no money. Big views. Good comments. No sales. And you will keep making more of it, because it felt like it worked. The fix isn’t better taste. It is numbers. That is the half most strategists skip, and it is the half a media buyer never skips.

What a media buyer actually does

Take a supplement brand selling on its own website. It has money set aside for ads, and somebody has to spend that money and make more back than they spent. That person is the media buyer. They own the ad account, set the budgets, decide which ads run and which get shut off, and they are judged on one thing: did the money come back with more money attached. Not views. Not engagement. Profit.

So a buyer thinks in steps and costs. What it cost to reach people. How many stayed past the opening. How many clicked. What each click cost. How many of those bought. What the order was worth. What was left after the product and the shipping. Every step has a number on it and a price. When sales drop, the buyer can point at the exact step where the money stopped moving.

Why brands added strategists

For years the buyer’s main skill was targeting. Then the platforms automated most of it. The machine got better at finding buyers than any human setting filters by hand, and targeting stopped being where the advantage was. Which left creative. The ad itself became the main thing that decides whether an account makes money.

That is a problem for a buyer, because a buyer can only move money between ads that already exist. When every ad in the account is tired, no amount of skill at the dashboard fixes it. So brands added a second seat. The buyer owns the money. The strategist owns what gets made. The buyer says this one is scaling, this one is dying, this angle cannot hit our target no matter what we spend. The strategist decides what the next batch should be and why.

On TikTok Shop, the machine holds the buyer’s seat

Now bring that to TikTok Shop, where there is no buyer. The brand sets a budget and a return target. GMV Max decides which videos get pushed and how hard. You don’t set a bid, pick an audience, or turn anything off. You supply content, and the system decides what it is worth. The data is thinner too: Seller Center gives you a smaller set of columns than a buyer on Meta would put up with.

So the buyer’s levers are gone. The buyer’s thinking is not. And since nobody is sitting in that seat handing you a read, you have to do that half yourself.

Two things carry over, and they are the useful ones.

• The funnel is still a funnel. Someone has to stop, stay, click, and buy. Those are separate steps and they fail separately. You get 2s view, 6s view, how far people watched, CTR, CTOR, and sales per ten thousand views. Smaller stack than Meta’s. Same shape. You can still name the step that broke.
• The arithmetic still holds. What a video earns per view is the click rate, times the rate at which clickers buy, times the order value. Same math a buyer runs.

Look at what is missing from that equation: attention. Nothing about how long people watched appears in it. A video can hold people beautifully and be worth nothing, because holding is not what gets paid. Clicks and purchases are. A strategist grading videos on watch time is chasing a number the machine is not paying for.

The example

Every video has four layers, and each one fails differently.

Most people work top down. Read the context, write the message, execute it, check the numbers at the end to see if they were right. Read it bottom up instead. Start with the behaviour and work backward to the layer that caused it.

That third row is where strategists give up too early. Some of it sits outside your video: price, reviews, the product page, how much people trust the brand. But a lot of it is yours. Did the hook pull in the wrong person, promising something this product doesn’t actually sell? Did you ever show the offer, or did you just talk about the product? Was there a reason to buy today instead of thinking about it? Did anyone else appear getting a result, or was it only your word? Did you tell people exactly what to tap and what happens after they tap it?

A video can sell the idea beautifully and never sell the purchase. That is message and execution, and you can fix all of it.

The numbers won’t tell you what to make

The buyer half gets you to the right step. It cannot get you any further. Your data can tell you the hook held. It cannot tell you the hook was a woman holding a bottle with tape over the label. It can tell you clicks were strong and sales were weak. It cannot tell you the strongest claim was buried twenty seconds in.

To close the loop you have to open the video and look. What was said. What was on screen while it was said. Was the product shown working. Was there real proof. That is the creator half, and no number replaces it.

When something works, squeeze it

The most common mistake after a win is going off to make something new. Don’t. A win means you found something the system will push. Now get everything out of it. Keep one thing, change one thing. Keep the angle, change the hook. Keep the hook, change the creator. Keep the whole middle, test three openings. Keep the argument, add a stronger demonstration. You are not asking what completely different thing to make tomorrow. You are asking where the signal is and how to point your next few slots at it.

Copying someone else’s winner depends on what you can see

Squeezing your own winner is safe, because you watched it happen. Copying another creator’s winner is the same move with less information. The problem isn’t copying. Copying is most of the job. The problem is copying blind.

When you pull a video off a public leaderboard, you are seeing one survivor. You are not seeing the other creators who ran nearly the same format that month and got nothing. Those videos exist. They just never appear on a board that only shows winners. So you cannot tell whether the format is why it won, or whether that creator caught a week when the brand was spending hard and the pool was still thin.

Change what you can see and the problem goes away. With the numbers for every creator running that brand, you can see the videos that used the same format and died. If eight ran it and one worked, that is a story about that creator or that week. If most of them did fine, the format is real and worth taking.

And once you can see that much, you can do better than copy. Look at the winner’s own funnel. A video can be the top earner on a brand and still leak. Maybe the hook is enormous and the click rate is only average. Maybe it clicks well and converts below the rest of the roster. That gap is your opening. Take the format, keep whatever is clearly carrying it, and fix the step it is losing at. That is how you beat a winning video instead of tying it.

Which is why you keep the record

Roster data solves this for one brand, in one window. Memory is what extends it. Write down the calls, and not just the wins. The angle you tried on three products that never sold. The hook that always tests well and never converts. The claim that pulls clicks and kills purchases. What the product and the payout looked like the week you ran it.

Do that and two things change. You stop re-testing dead ends you already paid for. And when a format looks strong on this brand this month, you can check it against every other time you have run something like it, instead of taking one good window as proof. Nobody remembers three hundred tests, and when a strategist leaves a brand, all of it walks out the door with them.

Do this in Verdict

Verdict is built around the four layers. Here is which part holds which.

1. 1
   Context. The Pre-Slot Screener and its Window Score: product stage, payout direction, how crowded the pool is, whether recent videos on the product are breaking out. It tells you whether the environment deserves a slot before you write anything.
2. 2
   Message. Waves groups scripts that share the same underlying argument, so you see the framework instead of one video. Script Lab is where you build your version off it.
3. 3
   Execution. The Element Library ranks pieces on the number each one owns, so hooks are judged on holding attention and calls to action are judged on clicks. Video Breakdown gives you the script and the on-screen action together.
4. 4
   Response. The Leak Finder shows one video, the full stack of numbers, and the step where it broke. Its creator drilldown runs that across a creator’s whole roster, so you can see which of their videos are worth a second try. Because it holds every creator on the brand, it also shows you where the current winner leaks, which is what you aim your version at.
5. 5
   Memory. Learnings holds every call with the environment it was made in, so the losing tests stay visible instead of disappearing.
6. 6
   And tomorrow's list, on Plan, is the output. Not a wall of charts to interpret. The next thing to make.