Most creators quit at post 6. The median first hit is post 5.

Ask ten people how many videos it takes before one goes viral and you will hear ten versions of the same belief: it is random, some people have it and some do not, and if your first handful did nothing, that is your answer. The data from 2.87 million creators says the opposite. The question has a measurable answer, and the answer is mostly about whether you kept going.

So how many posts does it actually take, what does "viral" even mean in the numbers, and what should a creator or a business do differently once they know?

First, define the thing

"Viral" is not a number anyone agrees on, so we used a strict one. Based on our analysis of 4.5 billion TikTok posts (2014 to 2026), a hit is a post in the top 1% by views within its country and week. The comparison group is posts from the same place at the same time, so a hit in a small market in a quiet week is still a hit. This is harder than "got more views than usual" and softer than "ten million views".

Among the 2.87 million creators in the dataset with at least 3 posts, we counted how many posts each one published before their first hit.

The median is 5. The spread is the real story.

The median number of posts before a creator's first top 1% hit is 5. Half of the creators who ever got a hit got it by their fifth post.

The quartiles tell you why the median is misleading on its own. The first quartile is 1: a quarter of creators who got a hit got it with their very first post. The third quartile is 16: a quarter needed more than sixteen posts. The 90th percentile is 38: one in ten creators who eventually got a hit needed thirty-eight or more posts to get there.

Now put that next to how people actually behave. Most accounts stop somewhere between post three and post eight. Our reading is that the most common exit point is right around the median, which means half of the people who would eventually have had a hit are quitting at the moment the odds were about to turn.

The probability by volume

Another way to see the same thing is the chance that an account has at least one hit, grouped by how many posts it has published:

• 3 to 4 posts: 1.6%
• 10 to 19 posts: 4.9%
• 20 to 49 posts: 10.9%
• 100 or more posts: 39%

Read that bottom line again. An account that reaches a hundred posts has close to a two in five chance of having produced a top 1% post. An account that stops at four has about one in sixty. The difference is not talent, because the dataset does not know anything about talent. The difference is twenty-five times more attempts.

Why volume works: each post is a separate test

In our own 100-reel experiment (one new account, 100 reels, 30 days, all produced by the pipeline that became Ultim), the median reel got 303 views, the best got 1,728, and the total was 48,254 views. None of them reached top 1%. But the shape of the data explained why volume matters.

TikTok gave each reel one pool of views in the first 12 to 24 hours and then the curve went flat. There was no slow build, no second wind, no "discovered a week later". Every reel was a single, independent test with a fast verdict.

That structure is why the creator data looks the way it does. If every post is a fresh draw, then the chance of at least one hit rises with the number of draws, and it rises in a way that rewards the people who are still posting at number fifty. Nothing about the first six results tells you much about the seventh, because the first six were six separate experiments, not one accumulating one.

Watch time did not open the gate

One more finding from our own 100-reel experiment that changes how to think about "almost viral": average watch time had a correlation of -0.11 with views. The reels people watched longest were not the ones that got shown to more people. So a creator who is improving retention post by post may be making better videos without making any progress on the signal that actually triggers distribution. That signal looks more like early sharing.

The early signal that actually predicts a hit

If you want to know within a day whether a post has a chance, the data points to shares, not comments.

Based on our analysis of 4.5 billion TikTok posts (2014 to 2026), a post with shares per view between 0.5% and 1% has a 2.5% chance of reaching the top 1%. At 2% shares per view or more, the chance is 5.6%. At zero shares, it is 0.01%. Saves per view at 2% or more give a 4.0% chance. Comments do not differentiate: a post with many comments is not meaningfully more likely to be a hit than one with few.

In our own 100-reel experiment, the 100 reels collected 12 shares in total. By this measure, none of them were ever in contention, and the data told us that on day one of each reel. The lesson for a creator is to stop counting likes and comments as progress and to watch the share rate on each post as the real early read.

Frequency: more is not better past a point

The obvious conclusion from "volume wins" is "post as much as possible". The data puts a ceiling on that.

Creators posting 1 to 3 times a week and creators posting 3 to 7 times a week show the same share of accounts with at least one hit. Above 7 posts a week, the chance per post drops. So the goal is not thirty posts a week. It is a sustained 3 to 7 a week for as many months as it takes to pass post fifty or post a hundred.

That reframes the problem from "how do I make one perfect video" to "how do I keep producing four decent videos a week for six months without burning out". For a solo creator, that is a discipline problem. For a business, it is a staffing problem, and it is usually the reason the account goes quiet after month one.

What four a day did in our test

In our own 100-reel experiment, moving from one reel a day to four reels a day in four different formats raised the daily total from about 800 to about 1,700 views. Views per reel fell, but less than proportionally. Each format seemed to get its own pool. This is also why format rotation matters more than repetition: four variations of the same idea is one test run four times, while four different formats is four tests.

What to change about each post while you keep going

Volume is the strategy, but each attempt should be a good one. The metadata levers from the same dataset are a 3x multiplier at best (a model on 6 million posts, AUC 0.74, found that the top 10% of posts by predicted score had about 3.5% chance of top 1% instead of 1%), so they are worth getting right on every post:

Length: 61 to 120 seconds carries a lift of 3.08; 8 to 15 seconds carries 0.63. Caption: 21 to 100 characters without hashtags, 1.7 to 1.8; empty, 0.27. Sound: own voice or own recording, 1.09; someone else's music, 0.79. Hook: the "part N" series is the only hook type gaining lift year over year (2.81 in 2024 to 3.01 in 2026), while most others lose 10 to 15% a year. Posting hour: 10:00 to 18:00 local, peak 15:00 to 16:00.

None of these turn a weak idea into a hit. All of them make each of your fifty tests a slightly better test.

The one percent

One last number to carry with you. One percent of creators make 77% of all hits. That sounds like proof that the game is rigged toward a few. It is also, in a dataset where volume is the strongest predictor we can measure, a description of the one percent who did not stop. The people who make most of the hits are overwhelmingly the people who have made most of the posts.

A business or a creator who wants to be in that group does not need a better first video. They need a system that gets them to post fifty, then a hundred, with 3 to 7 posts a week, each one a different test, each one judged by its share rate at 24 hours and not by how it felt to make.

Ultim was built to be that system: it writes, voices, edits and schedules the reels so that the only human job is approval and the only human decision is to keep going. The first thing to do, before any of that, is to see whether the hooks it writes from your website are worth fifty attempts. Paste the site into /tools/reel-hooks and judge the five it gives back.