How-To Guide

How do you read TikTok analytics to decide what to post next?

Read TikTok analytics in four steps: reach, hold, reaction and conversion, each against your own median, then change one thing in the next post.

10 min read

By the Superdirector editorial team. TikTok, Socialinsider, Rival IQ and vendor pages read November 7, 2026; Superdirector pricing read November 11, 2026.

At a glance

How do you read TikTok analytics to decide what to post next?

Read each TikTok video's analytics in four steps: reach (views and where they came from), hold (average watch time divided by video length), reaction (likes, comments, shares and saves divided by views) and conversion (new followers per 1,000 views). Compare each with the median of your last ten videos, find the earliest step where a video sits under half of it, and change one thing in the next post. TikTok itself says no single metric is necessarily more important than another, so judge trends across videos.

  • Reach, then hold, then reaction, then conversion: fix the earliest weak step first.
  • Compare with your own median of ten videos; benchmarks use different formulas (3.4% to 4.4%).
  • Read each video seven days after posting and keep your own log, since some metrics expire after seven days.
A numbered sequence from reach to hold to reaction to conversion, each labelled with the question it answers, and a stop rule that only the earliest step below half your median gets fixed.

The step order and the half-of-median stop rule are our method, not a TikTok rule. Original diagram; source basis: TikTok's analytics tutorial and the vendor guides read in November 2026.

Superdirector

Make it work for your brand

Find it. Remake it. Post it. Superdirector turns the viral formats in your niche into your brand's carousels, X posts and video scripts, and keeps the calendar full.

The four-step reading in one table

TikTok's own advice is to read analytics as trends. Its newsroom tutorial says no one engagement metric is necessarily more important than another, and that you should look at how audiences engage across your videos instead of one metric on one video (TikTok). That leaves a gap. A video has a dozen numbers, so which one decides your next post? We use four steps in a fixed order, because each step can only be fixed after the one before it.

StepQuestionNumbers to read in TikTok StudioWeak whenChange this first
1. ReachDid TikTok show it to people?Video views; traffic sources (For You, Following, profile, sounds, search, DMs)Views under half your medianTopic angle and packaging: first frame, on-screen text, caption
2. HoldDid people stay?Average watch time divided by video length; watched full videoUnder half your medianThe first three seconds, the length, the pacing
3. ReactionDid it move them?Likes, comments, shares and saves, divided by viewsUnder half your medianThe payoff and the ask: what to comment on or send
4. ConversionDid they want more?New followers per 1,000 views; profile viewsUnder half your medianBio, pinned videos, a promise of a series at the end

The median is the middle value of your last ten videos, so one viral post does not move it much. To find it by hand, sort the ten values and average the fifth and sixth. In a spreadsheet, =MEDIAN(D2:D11) does the same for a column of ten. The half-of-median line and the table itself are our method. TikTok publishes no such thresholds, and none of the guides we read gives a number for a good watch time. If a video is weak on more than one step, fix the earliest. That order is our reasoning, not a sourced rule: a viewer who leaves in the first seconds is unlikely to stay for the payoff or visit your profile, so fixing a later step first changes little.

Open TikTok Studio and pull your last ten videos

VEED and Buffer say analytics need a Creator or Business account, and that switching is free. Hootsuite's current guide also shows TikTok Studio analytics on a personal account, so check yours before you switch. On mobile, open your profile, tap TikTok Studio and then Analytics. On desktop, go to tiktok.com/tiktokstudio/analytics (Buffer, Sprout). In May 2024 TikTok announced that TikTok Studio replaces the older Creator tools, and said it reports account and content performance, audience behavior and monetization (TikTok).

Tab names depend on the account and the year. TikTok's 2020 tutorial lists Overview, Content and Followers. Sprout lists Overview, Content, Viewers, Followers and LIVE, and Hootsuite says a Business account shows Overview, Video, Audience and LIVE. Look for the metric names, not the tab names, and ignore guides that insist on one layout.

Four limits shape how you collect the numbers:

  • New data takes time. VEED says data starts to appear within 24 to 48 hours of switching and, later in the same guide, that it can take seven days. Check after 48 hours and treat the account as fully set up after seven days.
  • Some post metrics expire. Buffer says some individual post metrics are only available for the past seven days, and Metricool says reach or watch time can be blank for a video that has been inactive for more than seven days.
  • The Followers tab needs 100 followers before it shows data, according to Buffer.
  • History is one year. Hootsuite says TikTok offers 365 days of history, so a log you keep yourself is the only long record.

So write the numbers down. Read every video at the same age, seven days after posting, which fits those windows, and record the last ten in a sheet. TikTok's 2020 tutorial says you can download and export analytics when you use TikTok on desktop (TikTok), which saves retyping.

ColumnWhere it comes fromWhy you need it
Video lengthYour editWatch time means nothing until it is divided by length
ViewsContent tab or the video's analyticsStep 1, reach
Average watch time and watched full videoThe video's analyticsStep 2, hold
Likes, comments, shares, savesThe video's analyticsStep 3, reaction
Profile views and new followersOverview and the video's analyticsStep 4, conversion
Traffic sourcesThe video's analyticsWhether views came from strangers or followers

The log sheet has these columns, calculates each ratio against your median, and holds ten invented example rows to show how the maths works.

For one TikTok account, TikTok Studio and that sheet are enough. Pay for a tool only if you need TikTok next to Instagram or YouTube in one report, or more history than the 365 days Hootsuite says TikTok keeps. The reading below works on the same numbers wherever they come from.

Read reach: views and where they came from

Start with views against your own median, not against another account. Then open the traffic sources for the video. Hootsuite lists them as the For You feed, your profile, the Following feed, sounds, searches and DMs (Hootsuite). In one example VEED walks through, 96% of a video's views came from the For You page, which suggests most viewers found it through recommendation instead of through the Following feed. That is one video, not a norm.

Read the split this way. A video with most views from Following reached the audience you already have. A video with most views from For You was mostly recommended beyond your Following feed, so its later numbers say more about your content than about your existing followers. TikTok's tutorial says lots of factors determine how content is recommended in the For You feed (TikTok), so treat reach as a symptom, not a score you can aim at directly.

Low reach is the right thing to fix only when hold, reaction and conversion are healthy. A video with a strong watch time and few views has the content right and the packaging wrong: change the first frame, the on-screen text and the caption before you change the idea. The Content tab's list of videos trending on your account can help you spot your own best reach: TikTok's 2020 tutorial describes them as the top nine videos with the fastest view growth over the past seven days (TikTok).

Read hold: watch time against video length

TikTok's tutorial says video analytics show how long viewers watched, the total views and the average watch time (TikTok). Hootsuite adds two more: Watched full video is the percentage of viewers who watched in full, and video retention rate is the average percentage of the video that viewers watched (Hootsuite).

Seconds alone mislead. An average watch time of 6 seconds is a failure on a 30-second video and a success on a 7-second one. So divide average watch time by the video's length and compare the result with your median. In the invented example below, a trend remix held people for 4.5 seconds of 30, or 15%, against a median of 40%.

Compare videos of similar length where you can. TikTok's developer documentation groups video length as under 15 seconds, 15 to 60 seconds, 1 to 5 minutes and over 5 minutes (TikTok for Developers), and a 20-second video and a 4-minute video are different games. When hold is the weak step, Buffer suggests working on the first two to three seconds and tightening the edit, and our hooks guide covers openers.

The average hides where people leave. Buffer lists audience retention among the Content details a Business account shows. If your analytics show where viewers stop watching, read the shape of the drop before you change anything. This reading is ours, not TikTok's:

  • Most viewers gone in the first few seconds. The opening promised something the video did not start on. Rebuild the first line or the first frame and keep the rest.
  • A slow, even slide through the middle. The middle runs long. Cut pauses, repeated points and slow setup.
  • Steady until the last seconds, then a drop. That is people leaving when the video ends, which is normal. It only costs you if the payoff or the ask comes after the point where most have gone.
The same six seconds of average watch time is 86% of a 7-second video but 20% of a 30-second video, so divide watch time by length before comparing.

The same watch time divided by two different lengths. Original diagram with invented numbers; source basis: TikTok and Hootsuite descriptions of average watch time and watched full video.

Read reaction and conversion: shares, profile visits, follows

Reaction is engagement per view. Socialinsider defines engagement rate by views as the sum of likes, comments, shares and saves divided by views, times 100. TikTok's developer documentation lists a favorites count, which Hootsuite describes as the number of times a post has been saved to users' favorites. Calculate it per video, then compare it with your median. Buffer's tip is to watch shares, because they show what content people want to spread. Our own inference from that: a video with a high share rate is a good candidate for a follow-up or a series.

Benchmarks are a sanity check, not a target, and they do not use one formula. Rival IQ counts likes, comments and shares per view, while Socialinsider also counts saves, so the two sets below are not comparable with each other (Socialinsider, Rival IQ).

SourceDate and sampleFigureFormula
Socialinsider TikTok benchmarksBrand accounts of 1K to 1M followers, 40,694,424 videos, 2025 values4.40% (1-5K followers), 4.00% (5-10K), 3.90% (10-50K), 3.75% (50-100K), 3.95% (100K-1M)(likes + comments + shares + saves) / views
Rival IQ TikTok reportMay 2024, brand accounts3.4% average; about 22% of brands average under 2%, which Rival IQ reads as a sign that paid boosting supplies much of their reach(likes + comments + shares) / views

The Socialinsider figures are the 2025 column of its tier table. Both reports measure brand accounts, so use them as a rough range. If your engagement rate by views sits well under your own median, that matters more than whether it sits above a brand average.

Conversion is new followers per 1,000 views. It is the number a founder usually cares about most and the one most easily misread. VEED points out that people can follow from the For You page without visiting your profile (VEED), so new followers divided by profile views is not a clean conversion rate: some of those followers never saw the profile. Divide new followers by views instead, and read profile views as a sign of curiosity.

Name the weakest step, change one thing and log it

Here is the reading applied to ten invented videos. For each, compare each step's number with the median of the ten: 2,550 views, watch time at 40% of length, a 6.5% engagement rate by views and 2.1 new followers per 1,000 views. Each median is the average of the fifth and sixth values once the column is sorted: 2,500 and 2,600 views, 40% and 40% watch time, 6.4% and 6.6% engagement, 2.1 and 2.1 followers. A ratio is the video's number divided by the median, so V5's hold is 15% divided by 40%, or about 0.4 times.

VideoViewsWatch time / lengthEngagement rate by viewsFollowers per 1,000 viewsReading
V1 Three mistakes4,20040%6.1%2.1Reach 1.6x, the rest near the median: test a new first line
V2 Founder story1,90036%6.4%2.1Near the median; a 42-second story, try a shorter cut
V3 Screen demo2,60061%8.2%4.2Strong hold and follows at median reach: repackage for reach
V4 List of five3,10040%6.3%1.9Near the median on every step
V5 Trend remix12,80015%4.1%0.9Reach 5.0x, hold 0.4x, follows 0.4x: fix hold first
V6 How-to2,30050%12.5%7.8Reaction 1.9x and follows 3.7x: make more of these
V7 Opinion take1,50030%4.5%1.3Below median on every step, none under half: sharpen the claim
V8 Product demo2,80057%8.0%4.6Strong hold and follows: repackage for reach
V9 Customer story2,10037%6.6%1.4Near the median; follows 0.7x
V10 How-to, second take2,50054%12.4%8.0Confirms V6: the how-to format holds, reacts and converts

The trend remix, V5, has the most views by far and is the video most worth fixing. People came and left within seconds, and almost none followed. Hold is the earliest weak step (15% against 40%, about 0.4 times the median) and conversion is also under half (0.9 against 2.1, about 0.4 times), so the next post changes the first three seconds and keeps the topic. The two how-tos, V6 and V10, got ordinary reach but about twice the median reaction and three to four times the median follows. The next post is a how-to with the opening of V5, tested as a single change.

That last sentence is a hypothesis. Whether V5's opening would lift reach on a how-to is something the next ten videos will show, which is why you change one thing at a time and read the same four numbers again. Log each post as you go and keep the same weekday for the review. The log sheet holds the formulas and has a column for the first weak step and the one change you made. For a ratio definition behind the sheet, see the engagement rate glossary entry and the watch time entry.

The median needs some care in three situations. These rules are ours, not TikTok's:

  • Fewer than ten videos. Use the median of what you have once you have five, and mark the reading as provisional. Below five, post first and just log the numbers; a median of three videos moves too much to judge anything against.
  • Boosted or paid videos. Leave them out of the median and read them on their own. Paid promotion buys views, so it inflates reach and can pull the other ratios down. Rival IQ reads very low engagement by view as a sign that boosting supplies much of a brand's reach.
  • Every video looks the same. The median is relative, so an account that is weak across the board shows no video far under it. Compare the median of your latest ten with the median of the ten before, and check engagement against the benchmark range above. None of the guides we read gives a figure for good reach, so for views, your own trend is the only fair yardstick.
Four ratio bars for one invented video against a 1.0 median line: reach 5.0, hold 0.4, reaction 0.6 and conversion 0.4, with hold ringed as the first weak step.

One invented video read against the median of ten in the downloadable log. Original diagram with invented numbers; the basis is our half-of-median method, not TikTok's.

What we checked and what we did not

We read TikTok's own newsroom posts from 2020 and 2024 and its developer documentation, five vendor guides (VEED, Buffer, Hootsuite, Sprout Social and Metricool) and two benchmark reports (Socialinsider and Rival IQ) on November 7, 2026. Every sourced figure on this page names its source; the example numbers are ours.

We did not test inside a TikTok account, so tab names, where each number sits and whether a personal account shows analytics come from those guides and may look different on yours. Where the guides disagree, this page says so. TikTok's help center did not return article text to our fetcher, so TikTok's current metric definitions here come from its newsroom and developer pages and from the vendor guides.

The four-step order, the half-of-median line, the reading of where viewers leave and the ten example videos are ours. They are a way to decide what to change, not a TikTok rule, and the example numbers are invented.

Where Superdirector fits

Your TikTok Studio shows what your own videos did. It does not show which formats are working for other accounts in your niche, and that is one place the change in your next post can come from.

Superdirector, which publishes this guide, covers that part. Its main offer is $199 a month for one social account that its AI runs in the direction you set. It tracks accounts and keywords on TikTok, Instagram, YouTube and X, and hashtags on the first three, finds the formats taking off in your niche, and remakes their hook and structure as carousels, X posts and video scripts in your brand. On Growth it publishes eligible Instagram carousels and single X text posts by itself once you turn posting on for the account and answer one warning, with no per-post approval and nothing out in the first ten minutes. It does not post to TikTok, and AI video is upcoming, so filming and posting TikTok videos stay with you, and watch time, traffic sources and follower data stay in TikTok Studio. The pricing page lists two connected account slots on that plan; the service we run is one account, so confirm which account it will be before you pay. The $29 a month Starter plan is for hands-on work from Claude or ChatGPT and does not post. See what runs today.

Frequently asked questions

How do you read TikTok analytics to decide what to post next?

Read each video in four steps: reach (views and traffic source), hold (average watch time divided by video length), reaction (likes, comments, shares and saves divided by views) and conversion (new followers per 1,000 views). Compare each with the median of your last ten videos, fix the earliest step where a video falls under half of it, and change one thing in the next post. TikTok says no single metric is necessarily more important than another, so read trends across videos.

Where do you find analytics on TikTok?

On mobile, open your profile, tap TikTok Studio and then Analytics. On desktop go to tiktok.com/tiktokstudio/analytics. VEED and Buffer say analytics need a free Creator or Business account, while Hootsuite's current guide shows them on a personal account too, so check yours before switching. Tab names differ by account and date. After a switch, check for data after 48 hours and expect it to be complete within seven days. The Followers tab needs at least 100 followers.

What is a good engagement rate on TikTok?

Socialinsider's 2025 brand-account benchmarks run from 3.75% to 4.40% by follower tier, counting likes, comments, shares and saves over views. Rival IQ's May 2024 average engagement rate by views was 3.4%, counting likes, comments and shares. The formulas differ, so compare with your own median first.

What is a good average watch time on TikTok?

No source we read gives a threshold, and TikTok publishes none. Divide average watch time by the video's length and compare it with your last ten videos. A video under half your median is a signal to rework its first two to three seconds.

How often should you check TikTok analytics?

Weekly is enough. Read each video seven days after posting, because some post metrics are only available for seven days and Hootsuite reports that TikTok keeps 365 days of history, so keep your own log.

Related practical guides

Log your last ten videos today

Open the log sheet, enter views, watch time, reactions and followers for your last ten videos, and mark the earliest step where a video falls under half your median.