How TikTok Recommends Videos: Three Signals to Test

Creators often see unexpected changes in reach. A single dip does not prove an algorithm rollout or a punishment. Check the analytics for the specific posts, audience and format before changing your entire strategy.

Signal 1: How people respond to each video

TikTok identifies user interactions such as watching through or skipping a post, likes, shares and comments as factors in recommendations. These signals can be weighted differently for each person. There is no published blanket rule that a video must pass a follower-only test before non-followers can see it.

Experiment: Post two videos on the same subject with different openings. Compare average watch duration, completion rate, shares, comments and follows. Keep topic and length similar so the result is interpretable. Small samples are noisy; use several posts rather than treating one outcome as proof.

Signal 2: Retention matters, without a magic percentage

Completion and time spent watching can be useful measures of whether a particular video held interest, but TikTok does not publish a universal 70% completion requirement. A long tutorial and a short reaction clip should not be judged with a single unqualified percentage.

Experiment: Try a shorter introduction and a clear payoff. Check the average time watched and where viewers stop watching. If many leave before the useful content, revise the opening; if they stay for the explanation, preserve the depth. Do not delete helpful longer videos just to meet a made-up cutoff.

Signal 3: Match content to the right audience

TikTok says sounds, hashtags, viewer language, location and other content or user context can influence recommendations. Shares, saves and comments may be meaningful, but the service does not publish a universal ranked order that proves shares outweigh every other interaction.

Experiment: Write an accurate on-screen label and caption describing the real topic. Answer a specific question your audience has. Track search discovery and whether those viewers return. Avoid engagement bait, misleading hooks and manipulated engagement.

A simple four-week measurement plan

1. Baseline: Save metrics from five recent posts: subject, length, average watch time, completion, shares, and traffic source.
2. Test: Publish two consistent formats while changing one characteristic (opening, demonstration style, or caption).
3. Compare: Look at groups of posts, not one outlier. Segment search traffic from For You traffic when the analytics permit.
4. Decide: Repeat what attracts the intended audience and actually drives follows, repeat viewing or other legitimate goals.

Recheck TikTok's official recommendation explanation because platform guidance and features change. See our analytics and editing tools for workflows, not guaranteed growth.