Your First 10 TikTok Shop Videos: What to Make, in What Order

Almost everyone runs the first ten the same way: make one, post it, watch it do nothing, make a slightly different one, repeat. Ten videos later there is no pattern to read, because every video differed from the last in four ways at once.

The alternative is not more work. It is the same ten videos in a deliberate order, so that when you finish you know something.

The one rule the whole sequence rests on

One variable per batch, and read the batch, not the video.

Early distribution has a large random component. A video that gets nothing may be a bad video, or it may be an ordinary video that drew a bad ticket. You cannot tell from one, and you can barely tell from two. Three of the same kind, read together, is the smallest unit with signal in it.

That is why ten is a useful number: it is three batches with one spare, not ten attempts.

Batch 1 (videos 1–3): find out if anyone cares about the angle

Pick one product and one angle — the single reason you believe someone buys it. Not "it's good". Something like people think this is too small until they see it next to a hand, or nobody realises you can use it for X.

Make three videos that all argue that one thing, in different shapes. Different opening, different length, different footage. The angle is the constant; everything else varies.

What you are buying: whether that reason lands at all. If all three do nothing, the angle is wrong, and no amount of production polish saves a wrong angle.

⚠️ What you are not buying: which hook works. Three videos cannot answer that, and looking for it here is how people end up rebuilding their format on noise.

Batch 2 (videos 4–6): keep what worked, change the delivery

Now hold the angle constant — the one that got the best read, even if that read was modest — and vary how it is delivered. Faceless, hands-only, a person in frame, a transformation cut.

What you are buying: which format you can make work for this angle. If batch 2 clearly beats batch 1, the angle was fine and your delivery was the problem. If it does not move, the angle is weak and batch 3 should go back and try a different one rather than polishing further.

Most sellers skip this step and conclude from batch 1 that "TikTok doesn't work for my product". Usually they tested one angle in one delivery and called it a category verdict.

Batch 3 (videos 7–9): push the winner, or replace the angle

Two paths, and the batch-2 result chooses for you:

If something worked: make three more of it, deliberately similar. You are testing whether the result repeats, which is the only way to tell a real effect from a lucky draw. This is the least glamorous batch and the most informative.

If nothing worked: go back to angles. Different reason someone buys, best delivery you have found so far. Do not add production value — that is the expensive move that changes the fewest outcomes.

Video 10: the one you keep

The tenth is not an experiment. It is the best version of whatever survived, made carefully, and the one you actually put budget behind or pin to your profile.

If nothing survived, do not make a tenth. Spend it on a fourth batch instead — the information is worth more than the artefact at this stage.

Why ten and not thirty

The instinct is that more videos means more learning. It does not, past a point, and the reason is worth understanding because it also tells you when to stop reading and start scaling.

Ten videos in three batches gives you three comparisons. Thirty videos posted without structure gives you thirty anecdotes and still three comparisons at best, because the thing that produces a comparison is holding something still, not volume.

Volume becomes the right answer after you have an angle and a delivery that repeat. Then more is straightforwardly better, because you are no longer buying information — you are buying draws on a distribution you already know is favourable. Before that point, volume mostly buys you a larger pile of unreadable results.

⚠️ A concrete version of this: if you have posted thirty videos and cannot answer "which angle works for this product", you did not run thirty tests. You ran one test thirty times with the variables shuffled.

What "working" looks like at this stage

Set the bar where the data can actually reach it, or you will misread success as failure.

Do not judge by sales in the first ten. The number is usually zero and that is uninformative — the sample is too small for a conversion rate to mean anything.

Judge by whether the video was distributed at all. Did it get past the initial trickle? That is the question the first ten can answer, and it is upstream of everything else. A video nobody saw tells you nothing about your offer.

Then judge by comments that reveal intent. "Where do I get this" and "how did you make this" are worth more than a hundred passive views, because they show something landed hard enough to act on.

What to record, or the whole thing is wasted

This takes two minutes per video and it is the difference between ten videos and ten data points:

• The angle in one sentence, written before posting.
• What changed from the last video. If you cannot name it in one phrase, you changed too much.
• What you expected, roughly. Writing the expectation is what makes the result able to surprise you.
• The number, at a fixed age — 48 or 72 hours after posting, same for every video. Comparing a three-day-old video with a three-hour-old one is not a comparison.

The written expectation is the part people skip and the part that matters. Without it you will read whatever happened as confirmation of what you already believed, and ten videos will produce ten anecdotes.

Two failure modes worth naming

Stopping at two. Two poor results sit comfortably inside ordinary randomness. Almost every "this doesn't work for my niche" conclusion we see was drawn from a sample too small to support it.

Changing everything after each post. New hook, new length, new format, new time. It feels responsive and it destroys the experiment. If you must change several things, change them together as a new batch and read them as a batch.

Where the videos come from

Nothing above depends on how the videos get made — the sequence works with a phone. What changes with production speed is how long the ten take, and that matters more than it sounds: a plan that takes ten weeks gets abandoned; the same plan at three videos a week finishes.

Riffkit is built for that shape. You give it one video that already won, it models the emotion formula underneath, and rebuilds that formula as your own video with your product and your character — so a batch is three riffs of one structure rather than three shoots. You riff the formula, not the video.

Two related pieces if you are earlier or later than this: how to find winning TikTok videos worth modelling covers choosing the source, and faceless or on camera covers the delivery decision in batch 2.