PLAYBOOK

Reading TikTok analytics without lying to yourself

TikTok gives you more numbers than you can act on, and the ones it puts first — views, likes, followers — are the ones that tell you least about what to do next. Brand accounts stall because they optimize the number on the screen instead of the number underneath it.

This guide is about five metrics, in order of how much they matter, what each one means when it's low, and what to change. It ends with the monthly review that turns analytics from a mood into a plan.

How to read the numbers

Rates over totals — Views are a total; retention is a rate. Totals tell you how much distribution TikTok gave you, which you mostly don't control. Rates tell you what people did once they arrived, which you entirely control. Read the rates.
Formats over videos — One video is noise. The average of five videos in the same format is a signal. Tag every post with its format and compare formats, not posts.
Your baseline, not the internet's — 'Good retention is 60%' is someone else's account. Your median across the last twenty videos is the only baseline that means anything. Above it, the video worked; below it, it didn't.
Read monthly, glance weekly, never daily — Daily analytics produce daily mood swings and no decisions. Weekly, check that nothing broke. Monthly, sit down and change something.

The five numbers, in order

Three-second retention
The share of viewers still watching at second three. It measures exactly one thing: the hook. Nothing else about the video has happened yet. It's the first number to read because it gates every other number — a video with weak three-second retention never gets the chance to have a good completion rate.
Example
Median across your last twenty videos is 62%. A new video shows 48%: the hook failed, regardless of the rest. Another shows 74%: the hook worked — now read the next number.
Watch out
Fixing the script when the hook failed. If three-second retention is low, nothing after second three has been tested. Re-post with a different hook before touching anything else.
Average watch time and the retention curve
Where, exactly, people leave. The curve in TikTok Studio shows the second at which viewers drop; the shape tells you what went wrong. A cliff means a specific line lost them. A steady slope means the video was slightly too long everywhere.
Example
A cliff at second eleven. Watch second eleven: it's the sentence that starts with 'Our platform'. Cut it in the next five videos and the cliff disappears.
Watch out
Reading watch time without the curve. Eight seconds average watch time on a 30-second video could mean a great hook and a dead middle or a dull hook and a loyal core. The curve tells you which.
Completion rate
The share who watched to the end. It's TikTok's strongest quality signal for distribution, and it's the number that most rewards shorter videos. It also tells you whether the video kept the hook's promise — high three-second retention with low completion is a promise broken.
Example
Two videos, same format: 28 seconds at 41% completion, 19 seconds at 67%. The shorter one gets a second wave of distribution. Next month, that format is 20 seconds.
Watch out
Chasing completion by making every video eight seconds. Very short videos complete but rarely earn saves or follows. Completion is one of five numbers, not the only one.
Saves and shares per thousand views
Views are given to you; saves and shares are earned. Normalize them per thousand views so a small video and a big one are comparable. Saves mean the content was useful. Shares mean it was social. Both predict follower growth better than likes ever will.
Example
A tips video: 1,200 views, 38 saves — 32 per thousand. A skit: 40,000 views, 90 saves but 1,100 shares. Different jobs, both working. A video with 20,000 views and 4 saves and 6 shares is the one to worry about.
Watch out
Reading raw counts. 90 saves on 40,000 views is a worse save rate than 38 on 1,200. Divide before you compare.
Follower conversion and profile visits
How many viewers went to your profile and how many followed. This is the number that connects content to a business: it measures whether the video made someone want more of you, not just more of that video. Low across the board means the account has no recognizable point of view.
Example
A founder-note video: 3,000 views, 90 profile visits, 41 new followers. A trend remix: 60,000 views, 120 profile visits, 18 followers. The small one built the audience.
Watch out
Ignoring this because the numbers are small. They're always small. Compare them across formats and you'll find the one or two formats that actually grow the account — usually not the ones with the most views.
Traffic source: For You versus followers
The split between viewers who found you on For You and those who follow you. A high For You share means distribution; a high follower share means loyalty. Neither is bad. What matters is noticing when the ratio changes — a sudden drop in For You share usually means a format stopped working.
Example
Month one: 90% For You. Month four: 60% For You, 40% followers. The account has an audience now. If month five drops to 30% For You, the new content isn't reaching new people.
Watch out
Panicking at a low For You share on a video that did its job. A video for existing customers should have a low For You share. Read the source against the video's intended job.
Follower activity by hour
When your followers are actually online, from TikTok Studio, based on your account. This replaces every 'best time to post' chart on the internet the day you have four weeks of data. Post one hour before the peak so the video has early momentum when the audience arrives.
Example
Follower activity peaks at 21:00 on weekdays. Scheduled posting time moves to 20:00. Three-second retention doesn't change; first-hour views do, and so does distribution.
Watch out
Changing posting time weekly. The data needs consistent posting to be meaningful. Pick a time, hold it for a month, then adjust once.
What to ignore
Likes, total followers, and views on any single video. Likes are the cheapest action and correlate with nothing. Total followers is a lagging vanity number. Single-video views are distribution luck. None of them tell you what to change tomorrow.
Example
A video with 200,000 views, 9,000 likes and 3 saves taught you nothing except that a sound was trending. A video with 2,000 views and 60 saves told you what your audience wants more of.
Watch out
Reporting the ignorable numbers to your team or your boss because they're bigger. Report format-level retention, saves per thousand and follower conversion instead, and explain once why.

How brands misread their own data

Judging on day one — TikTok distributes in waves; a video can find its audience on day three or day ten. Read numbers after a week, and only compare videos that have had the same time.
Confusing reach with results — A viral video that brought zero followers and zero saves is a nice afternoon, not a strategy. Reach that doesn't convert to an audience is TikTok's win, not yours.
Changing three things at once — New hook, new format, new posting time in the same week — and now you don't know which one moved the number. One change per month per format.
Optimizing for the algorithm you imagine — The 'algorithm' is people, aggregated. It rewards videos people finish, save and share. Make those. Every hack that doesn't route through a real viewer's behaviour stops working within a quarter.

The monthly analytics review, in 30 minutes

Minutes 0–5Export or list the month's videos with format, length, three-second retention, completion, saves per thousand, shares per thousand and new followers. A spreadsheet is fine.
Minutes 5–15Average each number by format. Rank formats by completion and by saves per thousand. Note the format with the best follower conversion — it's rarely the one with the most views.
Minutes 15–22Open the retention curve on the three worst videos. Find the second where each one lost people. Write down the line or shot that caused it.
Minutes 22–28Decide one change per format: a shorter length, a cut line, a different hook pattern, a dropped format, an extra slot for the winner.
Minutes 28–30Write the three sentences you'd tell a teammate: what worked, what didn't, what changes next month. That paragraph is the whole report.

Analytics FAQ

What's a good retention rate on TikTok?: Better than your own median last month. Public benchmarks describe other niches, formats and audiences. After twenty videos your median is the only number that predicts anything about your next video.
How long should I wait before judging a video?: Seven days for a fair read, and compare only against videos that also had seven days. TikTok can re-surface a video weeks later; that's a bonus, not something to plan around.
Do I need a Business account for analytics?: A Business or Creator account gives you TikTok Studio with retention curves, follower activity and traffic sources. For a brand, yes — switch. The trade-off on commercial sounds is worth the data.
How do I report this to someone who only asks about followers?: Show three numbers per month: average completion by format, saves per thousand by format, and new followers by format. Then show which format you're doubling next month and why. Followers follow.

Analytics aren't a scoreboard. They're the only honest feedback you'll get about what to make next — if you read the rates, compare formats, and change one thing a month. Do that for six months and you'll know more about your audience than any chart on the internet does.

Want these written in your brand's voice?

Paste your website and Spoolt turns ideas like these into a month of scripted short-form — triage in Blitz, schedule to TikTok. Free to try, no card.

Start free →

How to read the numbers

Rates over totals — Views are a total; retention is a rate. Totals tell you how much distribution TikTok gave you, which you mostly don't control. Rates tell you what people did once they arrived, which you entirely control. Read the rates.

Formats over videos — One video is noise. The average of five videos in the same format is a signal. Tag every post with its format and compare formats, not posts.

Your baseline, not the internet's — 'Good retention is 60%' is someone else's account. Your median across the last twenty videos is the only baseline that means anything. Above it, the video worked; below it, it didn't.

Read monthly, glance weekly, never daily — Daily analytics produce daily mood swings and no decisions. Weekly, check that nothing broke. Monthly, sit down and change something.

The five numbers, in order

Three-second retention

The share of viewers still watching at second three. It measures exactly one thing: the hook. Nothing else about the video has happened yet. It's the first number to read because it gates every other number — a video with weak three-second retention never gets the chance to have a good completion rate.

Example

Median across your last twenty videos is 62%. A new video shows 48%: the hook failed, regardless of the rest. Another shows 74%: the hook worked — now read the next number.

Watch out

Fixing the script when the hook failed. If three-second retention is low, nothing after second three has been tested. Re-post with a different hook before touching anything else.

Average watch time and the retention curve

Where, exactly, people leave. The curve in TikTok Studio shows the second at which viewers drop; the shape tells you what went wrong. A cliff means a specific line lost them. A steady slope means the video was slightly too long everywhere.

Example

A cliff at second eleven. Watch second eleven: it's the sentence that starts with 'Our platform'. Cut it in the next five videos and the cliff disappears.

Watch out

Reading watch time without the curve. Eight seconds average watch time on a 30-second video could mean a great hook and a dead middle or a dull hook and a loyal core. The curve tells you which.

Completion rate

The share who watched to the end. It's TikTok's strongest quality signal for distribution, and it's the number that most rewards shorter videos. It also tells you whether the video kept the hook's promise — high three-second retention with low completion is a promise broken.

Example

Two videos, same format: 28 seconds at 41% completion, 19 seconds at 67%. The shorter one gets a second wave of distribution. Next month, that format is 20 seconds.

Watch out

Chasing completion by making every video eight seconds. Very short videos complete but rarely earn saves or follows. Completion is one of five numbers, not the only one.

Saves and shares per thousand views

Views are given to you; saves and shares are earned. Normalize them per thousand views so a small video and a big one are comparable. Saves mean the content was useful. Shares mean it was social. Both predict follower growth better than likes ever will.

Example

A tips video: 1,200 views, 38 saves — 32 per thousand. A skit: 40,000 views, 90 saves but 1,100 shares. Different jobs, both working. A video with 20,000 views and 4 saves and 6 shares is the one to worry about.

Watch out

Reading raw counts. 90 saves on 40,000 views is a worse save rate than 38 on 1,200. Divide before you compare.

Follower conversion and profile visits

How many viewers went to your profile and how many followed. This is the number that connects content to a business: it measures whether the video made someone want more of you, not just more of that video. Low across the board means the account has no recognizable point of view.

Example

A founder-note video: 3,000 views, 90 profile visits, 41 new followers. A trend remix: 60,000 views, 120 profile visits, 18 followers. The small one built the audience.

Watch out

Ignoring this because the numbers are small. They're always small. Compare them across formats and you'll find the one or two formats that actually grow the account — usually not the ones with the most views.

Traffic source: For You versus followers

The split between viewers who found you on For You and those who follow you. A high For You share means distribution; a high follower share means loyalty. Neither is bad. What matters is noticing when the ratio changes — a sudden drop in For You share usually means a format stopped working.

Example

Month one: 90% For You. Month four: 60% For You, 40% followers. The account has an audience now. If month five drops to 30% For You, the new content isn't reaching new people.

Watch out

Panicking at a low For You share on a video that did its job. A video for existing customers should have a low For You share. Read the source against the video's intended job.

Follower activity by hour

When your followers are actually online, from TikTok Studio, based on your account. This replaces every 'best time to post' chart on the internet the day you have four weeks of data. Post one hour before the peak so the video has early momentum when the audience arrives.

Example

Follower activity peaks at 21:00 on weekdays. Scheduled posting time moves to 20:00. Three-second retention doesn't change; first-hour views do, and so does distribution.

Watch out

Changing posting time weekly. The data needs consistent posting to be meaningful. Pick a time, hold it for a month, then adjust once.

What to ignore

Likes, total followers, and views on any single video. Likes are the cheapest action and correlate with nothing. Total followers is a lagging vanity number. Single-video views are distribution luck. None of them tell you what to change tomorrow.

Example

A video with 200,000 views, 9,000 likes and 3 saves taught you nothing except that a sound was trending. A video with 2,000 views and 60 saves told you what your audience wants more of.

Watch out

Reporting the ignorable numbers to your team or your boss because they're bigger. Report format-level retention, saves per thousand and follower conversion instead, and explain once why.

How brands misread their own data

Judging on day one — TikTok distributes in waves; a video can find its audience on day three or day ten. Read numbers after a week, and only compare videos that have had the same time.

Confusing reach with results — A viral video that brought zero followers and zero saves is a nice afternoon, not a strategy. Reach that doesn't convert to an audience is TikTok's win, not yours.

Changing three things at once — New hook, new format, new posting time in the same week — and now you don't know which one moved the number. One change per month per format.

Optimizing for the algorithm you imagine — The 'algorithm' is people, aggregated. It rewards videos people finish, save and share. Make those. Every hack that doesn't route through a real viewer's behaviour stops working within a quarter.

The monthly analytics review, in 30 minutes

Minutes 0–5Export or list the month's videos with format, length, three-second retention, completion, saves per thousand, shares per thousand and new followers. A spreadsheet is fine.

Minutes 5–15Average each number by format. Rank formats by completion and by saves per thousand. Note the format with the best follower conversion — it's rarely the one with the most views.

Minutes 15–22Open the retention curve on the three worst videos. Find the second where each one lost people. Write down the line or shot that caused it.

Minutes 22–28Decide one change per format: a shorter length, a cut line, a different hook pattern, a dropped format, an extra slot for the winner.

Minutes 28–30Write the three sentences you'd tell a teammate: what worked, what didn't, what changes next month. That paragraph is the whole report.

Analytics FAQ

What's a good retention rate on TikTok?

Better than your own median last month. Public benchmarks describe other niches, formats and audiences. After twenty videos your median is the only number that predicts anything about your next video.

How long should I wait before judging a video?

Seven days for a fair read, and compare only against videos that also had seven days. TikTok can re-surface a video weeks later; that's a bonus, not something to plan around.

Do I need a Business account for analytics?

A Business or Creator account gives you TikTok Studio with retention curves, follower activity and traffic sources. For a brand, yes — switch. The trade-off on commercial sounds is worth the data.

How do I report this to someone who only asks about followers?

Show three numbers per month: average completion by format, saves per thousand by format, and new followers by format. Then show which format you're doubling next month and why. Followers follow.