TikTok Growth Basics for New Accounts

TikTok is the platform where a brand-new account with no followers can reach a million people, and where an account with a million followers can post something that reaches four thousand. Both facts come from the same design decision, and it is the one thing to understand before anything else.

How distribution works

When you publish, TikTok shows the video to a small sample and measures the response. If the signals are strong it widens the audience; if not, it stops. TikTok describes the general mechanism in its recommendation documentation — user interactions, video information and device settings, without published weightings.

What is observable is that the interaction signals dominate, and that the process repeats at each stage. A video that keeps performing keeps expanding, which is why TikTok reach distributions are so extreme compared with other platforms.

The consequence for a new account: you are not disadvantaged in the way you would be on a platform where followers gate distribution. The test audience is not much smaller for a new account than for an established one. This is genuinely unusual and it is the platform’s main attraction for people starting from nothing.

The signals that decide it

In rough order of weight, inferred from behaviour and from what TikTok states:

1. Completion rate. The proportion who watched to the end. The single strongest input.
2. Rewatches. Watch time exceeding video length. Unambiguous and heavily weighted.
3. Shares. Particularly shares to individuals, which cost the sharer something.
4. Comments, weighted for substance rather than count.
5. Saves. Strong for reference-type content.
6. Likes. The weakest countable signal, and correspondingly the cheapest to buy.
7. Follows from the video. Matters for the account rather than the individual post.

Completion being dominant explains most of what looks like folklore about the platform. Very short videos over-perform because completion is easier to achieve, not because brevity is inherently better. Loops work because a seamless restart converts into a rewatch. Hooks matter because everything downstream is conditional on not being scrolled past. Which signals matter covers the general principle across platforms.

What this means practically

The first second

More decisive here than anywhere else. A viewer who scrolls at 0.4 seconds produces a failed completion, and there is no later moment that can recover a signal never generated.

Practically: start with the thing itself. Introductions, logo animations and “hey guys” openings are all instructions to scroll.

Length

Shorter is easier to complete, so shorter has a structural advantage — but a longer video that holds people generates far more watch time per view, which is also weighted. The rule that follows is not “be short” but “be exactly as long as the material justifies”, with nothing padded.

Subject coherence

Because the system chooses a test audience based on what it has learned to associate with you, an incoherent account gets tested on the wrong people every time. On TikTok this is more punishing than elsewhere, since a bad test audience means a bad first sample, which means the post stops immediately.

Follower count

Barely relevant to distribution. Your followers are one audience segment among several the system may test, and a video that performs badly with them will not be shown further regardless of how many there are.

Which makes buying TikTok followers unusually pointless — even by the standards of this category. The mechanism it is normally justified by does not exist here, and the human social-proof effect is weaker on a platform where people consume in a feed and rarely visit profiles. Social proof covers where that effect actually operates.

Starting an account

1. Pick a narrow subject. Narrower than feels comfortable. The system has to learn who to show you to, and it learns faster from consistency.
2. Post frequently at the start. Each post is an independent trial, and more trials means faster learning — both yours and the system’s.
3. Watch the retention graph. TikTok shows where viewers dropped off. That graph is the most direct feedback available on any platform.
4. Repeat what worked, structurally. Not the same video — the same opening structure, pacing, or format.
5. Expect high variance. On TikTok more than anywhere, most posts do little and occasionally one does a great deal. That distribution is the platform working normally.

What does not work

• Buying followers. Distribution is per video and follower count is a minor input. There is not even a plausible mechanism to argue about here.
• Buying views. Purchased views arrive with negligible watch time, which drags down the completion rate that is the dominant ranking signal. This is actively counterproductive rather than merely useless.
• Deleting underperforming videos. A common belief with no stated mechanism behind it. Distribution is per video; a video that did nothing is not holding anything back.
• Hashtag volume. Content signals matter, but stuffing thirty tags does not compensate for a video nobody finished.

Reading your own analytics

TikTok exposes more useful per-video data than most platforms, and three numbers carry almost all of it.

Average watch time against video length. Divide one by the other for an approximate completion rate. This is the number to optimise and the only one that reliably predicts whether a video will keep spreading.

Traffic source. TikTok reports the share of views from the For You feed versus from your profile, from search, and from following. A high For You share means the system is testing you widely, which is the leading indicator for growth. A video that got views mainly from your profile did not spread.

The retention curve. Where viewers left. Like YouTube’s, it identifies the exact second that lost people, and the fix is almost always to delete everything before it.

What to do with this: compare your best video against your worst on all three, and look for the structural difference rather than the topical one. It is usually the opening, and it is usually fixable.

What high variance means for how you work

The reach distribution on TikTok is extreme — most posts do modestly and occasionally one does enormously. That is not a sign of doing something wrong; it is the shape of the system.

Two consequences for how to operate:

Judge in batches, not individually. A single video’s performance is dominated by noise. Ten videos in a format tell you whether the format works; one tells you nothing, and reacting to one is how people end up changing direction every week.

Do not over-learn from the outlier. A video that reached two million did so partly because of what it was and partly because of where the sampling landed. Copy its structure, not its specifics, and expect the copy to do a fraction of the numbers.

This is also why deleting underperformers is counterproductive beyond the missing-data problem: in a high-variance system, most posts are supposed to be unremarkable.

Where TikTok growth is genuinely hard

The flip side of per-video distribution: nothing accumulates. On YouTube a back catalogue keeps earning views for years. On TikTok, reach is overwhelmingly recent, so an account that stops posting stops existing, and the audience is attached to the feed rather than to you.

Which is the argument for treating TikTok as an acquisition channel that points somewhere durable — a mailing list, a channel, a site you own — rather than as the destination. Organic vs paid growth covers how the channels fit together, and the social media growth guide covers what the platforms have in common.