How to find winning formats on Instagram and TikTok: test, tweak, and double down

Which reels work for your account is trial and error. Here is a routine for it: test formats across several reels, read the numbers that matter, call a winner, and remake it without posting the same reel twice.

By Ricardo Rivero

Scaling consumer apps on TikTok and Instagram.

November 14, 2026

Over-the-shoulder view of a bakery owner pinning printed video frames in columns on a cork board in the back room

Nobody can tell you in advance which reels will work for your account. You find out by trial and error: you test formats, tweak them, adjust from the numbers, and when one format wins, you double down on it. Then you keep testing a little, because the next winner is somewhere in the tests.

This guide turns that into a routine: what a format is, how to test one across several reels, which numbers to read, when to call a winner, and how to remake a winner without copying it. It borrows two rules from TikTok's own guide to testing ads, and it ends with a real Hook Studio session where one format is remade three ways. Hook Studio is an AI reel maker for TikTok and Instagram. It is not affiliated with TikTok, Instagram, or Meta.

The short answer

A format is the recipe of a reel: who is on screen, how it opens, how it is shot, how long it runs, and how it ends. Test two or three formats that differ in kind, with three reels each on three different ideas. Read every reel at the same age, compare the middle reel of each format instead of its best one, and use rates such as shares per 100 views instead of raw views. A format that beats your usual numbers on most of its reels is a winner. Make most of your next reels in it, with a new hook, person, or setting each time, and keep a smaller share of your posts for new tests.

A format is a recipe, not a post

A post is one reel, one idea, one day. A format is the structure you can fill again with a new idea tomorrow. When people say "that format went viral", they mean the recipe, not the video.

Part of the format	Examples
Who is on screen	You talking to the camera, a two-person skit, hands only, no face at all
How it opens	A question, the result first, a bold claim, a mistake
How it is shot	A phone in a car mount, a POV, text over footage, a before and after
How long it runs	7 seconds, 15 seconds, 30 seconds
How it ends	A punchline, a reveal, a question, a call to action

For 12 formats with their timing by the second, see our reel templates. For openings, see the 50 hook examples.

Why one reel tells you almost nothing

TikTok says neither your follower count nor your past hits is a direct factor in what it recommends, and views are not spread evenly: a few videos get far more than the rest, and the same account has both. We go through what TikTok has published in how to go viral on TikTok. So one reel in a format is one ticket. A flop can be a bad day, and a hit can be a lucky idea. Three reels in the same format, on three different ideas, start to tell you about the format itself.

How to test a format: the loop

The loop in one line

Two or three formats, three reels each, one variable at a time. Read every reel at the same age, compare the middle reels, keep a log, tweak the best format, and test again.

1Pick two or three formats that differ in kind. A talking head against the same talking head with a new background teaches you nothing. Talking to the camera, a two-person skit, and text over footage do.
2Make three reels in each format, on three different ideas. If every reel in a format carries the same idea, you are testing the idea. Different ideas let the format show through.
3Hold everything else steady. Similar length, the same posting hour, no paid promotion on any test. If you change the posting time halfway, the test cannot tell the format from the clock.
4On Instagram, post the tests as trial reels. Trials go to people who do not follow you first, so your followers do not see the losers. The rules and the steps are in our guide to trial reels.
5Read every reel at the same age. 48 hours after posting for an Instagram trial, as in our 3-hook trial. Day 3 on TikTok, because TikTok says its analytics can take 1 to 2 days to update.
6Compare the middle reel of each format, not its best one. One hit drags an average far up. With three reels, the middle value is the honest one.
7Keep a log. One line per reel: the format, the idea, the hook, the length, and the numbers. After a few rounds the log, not your memory, tells you what works.
8Tweak the best format one thing at a time, then test again. A new hook, a new length, a new ending. Change one and keep the rest.

The last two steps come straight from TikTok's guide to ad testing, written for advertisers but true for any post. On one variable: "Change only one element per test, such as video length, headline wording, or call to action. Testing multiple variables at once makes it unclear what drove performance differences." On the log: "Document what works, what doesn't, and why."

Which numbers to read

Read rates, not counts. Views swing with timing and luck, so never judge a format on views alone. A reel with fewer views but more shares per 100 views did better with the people who saw it.

App	Number	What it tells you
Instagram	Skip rate	The share of views where people skipped in the first 3 seconds. It grades the opening.
Instagram	Average watch time	Whether people stay. Instagram's head, Adam Mosseri, named watch time one of the three top ranking signals in January 2025.
Instagram	Sends per reach and likes per reach	The other two of those signals. Mosseri said sends count slightly more for reaching people who do not follow you yet.
TikTok	Average watch time and watched full video	TikTok gives finishing a longer video as an example of a strong signal of interest.
TikTok	Shares and follows	Likes, shares, and follows are among the interactions TikTok says it uses.
Both	Shares per 100 views	A rate you can compare across reels with very different views.

Pick the number that matches what you want from the format. To reach new people, watch sends and shares. To grow, watch follows. To know if the format holds attention at all, watch average watch time first. The sources for the Instagram signals are in how to grow on Instagram.

When to call a winner

A winner beats your usual numbers on most of its reels, not on one. Your usual is the middle value of your last 10 reels on the number you picked. Compare against your own account, never against someone else's screenshot. And wait for the numbers to settle. TikTok's guide to ad testing puts it plainly: "Stopping too early risks drawing false conclusions. Wait until performance metrics stabilize before calling a winner."

What you see	What it means	Next step
One format beats your usual on 2 of its 3 reels	A winner	Double down: most of your next reels in this format.
One reel took off, the other two did not	A lead, not a winner. The idea or luck may have carried it.	Make two more reels in the format before you decide.
All the formats are close	They were too alike	Test formats that differ in kind, not in details.
Each new reel in a winning format does worse than the last	Your remakes are too close to each other	Change more than the words: a new person, setting, or idea.
Every reel has very few views	No result, not a loss	Check that the account is in good standing, then run the test again.

Double down: most of your next reels in the winner

Once a format wins, it earns most of your posts. Our rule: spend 60 to 80% of your budget on remaking winners, and 20 to 40% on testing new formats and new video models. Each test still gets several reels, never one. The math, with real credit prices, is in the 70/30 rule.

Remake the winner, never repost it

Doubling down does not mean posting the same reel again. Instagram says a trial reel may get limited reach if it detects that you have shared the same content before, and it does not recommend accounts that mostly repost other people's content. Keep the format and change what goes in it.

Change	Example	What it tests
The hook	A new first frame and a new first line	Whether the opening or the body carries the format
The idea	A new tip, a new story, a new product	Whether the format works beyond its first topic
The person	A teammate, a customer, another face	Whether the format depends on one face
The setting	From the car to the porch, from the kitchen to the street	Whether the place is part of the recipe

While you learn what carries a format, change one of these per remake. Once you know, change two or three at a time, so your feed does not look like the same reel on repeat.

A real example: one format, three remakes

We made this in Hook Studio on November 14, 2026, as stills only. The format is a realtor's quick tip: one person, one tip, said in one breath straight to a phone in a mount. Grant and Valentina are AI people from Hook Studio's premade cast, and the tip is made up for the demo. We kept the stills the realism check failed, because they show what to fix before a test goes out.

Grant, a man with short dark curls in a white tee, in the driver's seat of a parked car, talking to a phone in a dashboard mount — The format: one tip, said in one breath to a phone in a dashboard mount. The realism check failed this still: the sky shows in the top left corner, as if the camera sat outside the car, and the visor label has garbled print. It shows the format; we would remake it before posting.

Grant in the same parked car, holding a house key up toward the phone with a half smile — New hook: the same person and car, with a new first frame, a house key held up to the lens, and a new first line. The check flagged the visor label again, and the key came out smaller than we asked.

Valentina, a woman with long brunette hair in a cream knit top, in the driver's seat of her parked car, talking to a phone in a mount — New person: Valentina in her own parked car, the same framing and the same line. It passed the check.

Grant standing on the front porch of a gray house with a white door, talking to the camera — New setting: Grant on the porch of a house for sale, with the same line. It passed the check, with one note: the camera sits in the yard, not on the porch railing as we asked.

Reel	What changed	First line
The original	Nothing	"One question to ask before you list your house: how did you get that price?"
New hook	The first frame and the first line	"Before you hand anyone this key, ask one question."
New person	The person	The same line as the original
New setting	The place	The same line as the original

Each remake changes one thing. Post them as a test, and the numbers say which change helped: the opening, the face, or the place.
Stills first. The four stills cost 41 credits in all. You pay for motion only for the remakes you will post: a 5 second talking clip on Kling 3 is 143 credits in Hook Studio.
Fix what repeats. Both car stills came back with a garbled label on the sun visor. When the same flaw shows up twice, name the fix in the prompt of every remake, here "plain visors, no labels".

Where Hook Studio fits

Describe the winning format in the chat once, and ask for the next remakes: a new hook, a new person, a new setting. You see every remake as a still before you pay for a clip, and every clip waits for your approval on a card with its price. To start from a reel you like, paste its link: importing it costs 5 credits, and the agent sees its frames and caption, so you can say "same structure, my product". Save the remakes you plan on the idea board, and the next round is ready when the test ends.

More on testing: the weekend A/B test of 20 TikTok carousels, why two or three variants are often too few, and how to remix your winners.

Questions about winning formats

What is a winning format?

A reel structure that beats your usual numbers on most of the reels you make in it, not on one. The structure is who is on screen, how the reel opens, how it is shot, how long it runs, and how it ends. A single viral reel is a lead until two more reels in the same format do well.

How many reels do I need to test a format?

At least three per format, each on a different idea. One reel can win or lose on luck or on its idea. Three reels on three ideas show you the format.

How long should I wait before I judge a reel?

Read every reel at the same age. We use 48 hours for Instagram trial reels and day 3 on TikTok, because TikTok says its analytics can take 1 to 2 days to update. TikTok's ad testing guide also warns that stopping too early risks false conclusions.

Which number decides the winner?

The one that matches your goal, read as a rate. Skip rate grades the opening on Instagram, average watch time shows whether people stay, sends and shares show reach to new people, and follows show growth. Never judge a format on views alone.

Can I repost a winning reel?

Remake it instead. Instagram says a trial reel may get limited reach if it detects that you have shared the same content before. Keep the format and change the hook, the idea, the person, or the setting.

How much of my content should be tests?

Our rule is 20 to 40% of your budget on tests of new formats and video models, and 60 to 80% on remaking winners. Each test gets several reels. The full split is in our guide to the 70/30 rule.

Does this work with AI video?

Yes, and the video model becomes one more thing you can test. Keep the format and the line the same and change only the model, so the numbers tell you about the model. Make the stills first: they are the cheap part.

Remake your winning format

Describe the format once, see every remake as a still, and pay for motion only for the ones you post.

Make your first reel