TikTok Metrics Dictionary: Views, Reach, Watch Time, Completion, Shares, Saves and Followers

TikTok analytics metrics are counts, durations and rates that describe exposure, viewing behavior, interactions and audience change. A number only becomes useful when you know its source, scope, denominator and time window. Ten thousand views usually means ten thousand recorded plays, not ten thousand people. A completion rate is a share of views, not a raw total.

TikTok Metrics Explained: Source, Scope, Denominator and Time

TikTok analytics does not have one universal reporting screen. Creators may see data in TikTok Studio on mobile or web. Business Accounts can also use Web Business Suite. Authorized brand collaborations may appear in TikTok One. Paid campaigns have separate definitions in Ads Manager. APIs and third-party reporting tools expose another set of field names.

Those surfaces overlap, but they are not interchangeable. TikTok's current Studio documentation separates account analytics into Overview, Content, Viewers and Followers. It also provides post-level analytics. The fields shown can differ by account, device, region and product version. TikTok's Studio documentation describes the available sections, while TikTok's creator-tools documentation names total views, net followers and likes among the headline metrics.

Business Accounts have another reporting surface. TikTok's Web Business Suite documentation lists downloadable Reach, Engagement, Conversion and Followers data plus a separate Video view. It supports fixed ranges such as 7, 28 and 60 days as well as custom dates. TikTok One uses terms such as total video views, unique viewers, total play time, average view time, video completion rate and favorites. Its project-reporting metric definitions are useful, but they describe TikTok One rather than every TikTok Studio account.

Before interpreting a number, record four things:

This dictionary focuses on organic creator and account measurement. Watch-time and completion definitions apply to video posts. Photo posts and LIVE have separate analytics. Paid metrics are mentioned only when the distinction prevents a reporting error.

TikTok Metrics Dictionary

Quick reference table

Views

Definition and source: Views are the number of video views or plays recorded for a post. At account level, post views combine views across the content included in the selected period. TikTok's Research API defines view_count as the total number of views a video has received. It does not define that count as unique people. TikTok's Research API codebook also treats a public video's view count as a cumulative public field.

Native or calculated: Native.

Unit and scope: A count. It can be post-level, account-level or project-level. A public post count is generally a current cumulative value, while an account dashboard may report only views recorded inside a selected date range.

Denominator: None. Views become a denominator when you calculate rates such as view-based engagement or completion.

Best comparison window: Compare posts at the same age, such as each post's first 24 hours or first 7 days. For account reporting, use equal calendar windows.

What it can diagnose: How much recorded exposure a post or account received.

What it cannot prove: Views do not prove how many different people watched, how long they watched or what they did next. TikTok does not publish a complete organic counting specification that lets creators reconstruct every view from raw events.

Do not borrow the Ads Manager definition and apply it to organic views. In Ads Manager, TikTok defines a video view as a video starting to play within an impression session and documents separate rules for replays and repeat impression sessions. Those are paid-reporting rules from TikTok's video play metrics documentation, not a universal definition for TikTok Studio.

Reach, unique viewers and reached audience

Definition and source: This family of metrics measures distinct audience rather than total plays. TikTok's Organic API exposes a reach field for authorized Business Account video reporting. TikTok One uses Unique viewers and defines it as the number of unique viewers who watched a video or at least one video in a project. TikTok Studio uses a Viewers area and some product versions may show labels such as Total viewers or Reached audience.

There is no safe reason to rename every one of these fields as reach. Keep the exact label from the screen or export. Treat it as unique only when that reporting surface describes it that way. The current field set in TikTok's official Business API collection includes video views, reach, total time watched, average time watched and full-video watched rate. TikTok's official Business API request reference shows those fields together.

Native or calculated: Native where available.

Unit and scope: A deduplicated count of accounts or viewers within the stated post, account or project scope and time window.

Denominator: None. A useful calculated companion is:

Views per unique viewer = Views ÷ Unique viewers

Both values must come from the same surface, scope and window.

Best comparison window: Use a fixed post-age window or the same account date range. Do not add post-level unique viewers to estimate account-level unique viewers because the same person may appear under several posts.

What it can diagnose: Whether exposure came from a broad audience or repeated viewing by a smaller audience.

What it cannot prove: A unique viewer count does not show attention depth or future loyalty. Paid reach also has its own methodology. TikTok says Ads Manager reach can use sampling and estimation, so its paid reach definition should not be silently combined with organic reach.

Total watch time

Definition and source: Total watch time, also labelled Total play time or exposed as total_time_watched, is the cumulative time spent watching the content in the stated scope. TikTok One defines Total Play Time as the cumulative amount of time users spend watching a project or video. The Organic API exposes the same concept for supported Business Account videos.

Native or calculated: Native on supported post, project and API reports.

Unit and scope: A duration, usually displayed in seconds, minutes or hours. It may describe one video or an aggregate project. Record the unit before exporting or calculating.

Denominator: None. It is a total.

Best comparison window: Match post age and compare videos within similar length cohorts. A 60-second video has more possible watch time per play than a 10-second video.

What it can diagnose: The total volume of recorded viewing generated by a post or content group.

What it cannot prove: It does not describe a typical view. A high total may come from more views, longer videos, longer viewing sessions or a combination of them. TikTok's public organic documentation does not provide enough event-level detail to infer replay behavior from total watch time alone.

Average watch time

Definition and source: Average watch time is the average amount of time watched per view on the relevant organic reporting surfaces. TikTok One calls it Average view time and defines it as total play time divided by the number of views. The Organic API uses average_time_watched.

Native or calculated: Native on supported surfaces and derivable when total watch time and views use the same scope.

Average watch time = Total watch time ÷ Video views

Unit and scope: Usually seconds per view for one video. A project-level average may weight videos by their views, so it is not necessarily the simple average of the displayed post averages.

Denominator: Video views, not unique viewers, for the TikTok One organic definition.

Best comparison window: Use the same post-age window and similar video lengths. Always record video duration beside average watch time.

What it can diagnose: Whether two comparable videos produced different average viewing depth despite similar view counts.

What it cannot prove: An average hides the shape of the underlying behavior. An average watch time of 15 seconds could come from many viewers leaving near 15 seconds or from a mix of very short and complete views. It also cannot be compared directly with Ads Manager's Average play time per user, which uses users rather than views as its denominator.

Completion rate or full-video watched rate

Definition and source: Completion rate is the percentage of views that reached the end of the video. TikTok One displays Video completion rate and documents the formula as complete video views divided by total video views. The Organic API exposes full_video_watched_rate. TikTok Studio versions may use Watched full video, a finish-rate label or no visible completion field.

Native or calculated: Native on TikTok One and supported Organic API or Studio reports. If your surface does not display it but provides complete views and total views, calculate it transparently.

Completion rate (%) = Complete video views ÷ Total video views × 100

Unit and scope: A percentage for one video or a defined group of videos.

Denominator: Total video views on the same surface and in the same window. TikTok One's explanatory copy refers to viewers, but its published formula uses video views. That distinction matters because views are not necessarily unique people.

Best comparison window: Match post age and use tight video-length cohorts. Completion on a 10-second clip and a 90-second explanation answers different viewing questions.

What it can diagnose: How often a recorded view reached the video's end.

What it cannot prove: Completion rate does not reveal where incomplete views ended, how many unique people completed the video or whether completion caused more distribution. TikTok has not published a fixed algorithm score that creators can derive from this rate.

Likes and comments

Definition and source: A like is a heart action recorded on a TikTok video. A comment count is the number of comments posted on the video. Both are native post counts. TikTok's Research API codebook defines like count, comment count, share count and view count as separate public video fields.

Native or calculated: Native.

Unit and scope: Counts at post level. Account and project reports may aggregate them across included videos.

Denominator: None until you turn them into rates. A like rate by views is Likes ÷ Views × 100. A comment rate by views follows the same structure.

Best comparison window: Use the same post-age window and comparable traffic conditions.

What they can diagnose: The volume and type of visible interaction. Comments can mark conversation volume while likes record a lighter response.

What they cannot prove: A count does not reveal motive, sentiment or quality. Comments can be positive, negative, repetitive or unrelated. Likes and comments can also change after moderation, deletion or privacy changes.

Shares

Definition and source: Shares are the total number of times TikTok records a video being shared through its Share action. TikTok's codebook describes share_count as the number of times a video has been shared by clicking the Share button. TikTok One reports total, organic and paid shares for supported projects.

Native or calculated: Native.

Unit and scope: A post interaction count or an aggregate across a stated set of videos.

Denominator: None. A calculated share rate by views is:

Share rate (%) = Shares ÷ Views × 100

Best comparison window: Match post age, content cohort and traffic source.

What it can diagnose: How often viewers took a share action relative to comparable posts.

What it cannot prove: Share count does not disclose every destination, recipient, resulting view or incremental reach. A share is an action, not a guaranteed new viewer.

Saves or Favorites

Definition and source: TikTok's current first-party reporting term is Favorites. Marketers often call the same action a save. TikTok One defines Favorites as the total number of favorites received by videos in a project. TikTok's Research API uses favorites_count, defined as the number of favorites a video receives.

TikTok added favorites_count to several Research Tools video endpoints on May 21, 2026, according to the TikTok for Developers changelog. That confirms the current field, but it does not mean every creator analytics screen or historical export now displays it.

Native or calculated: Native where supported.

Unit and scope: A post count or a project aggregate.

Denominator: None. A calculated favorite rate by views is:

Favorite rate (%) = Favorites ÷ Views × 100

Best comparison window: Match post age and compare similar content formats or topics.

What it can diagnose: How often a favorite action occurred. It can be useful when comparing content designed to be revisited, but the metric itself does not reveal the viewer's reason.

What it cannot prove: Favorites do not prove a later return, a purchase or wider distribution. They are not the same as downloading a video, favoriting a TikTok Shop product or adding an advertisement to a Creative Center collection.

Followers

Definition and source: Follower count is the number of accounts following a TikTok profile at a given point. TikTok's User Info API exposes follower_count as a current profile statistic. TikTok's Get User Info documentation distinguishes it from following count, total likes received and public video count.

TikTok Studio also reports Net followers in its account analytics. Web Business Suite provides new follower data and follower reporting for selected periods.

Native or calculated: Current followers, new followers and net followers are native where shown.

Unit and scope: An account-level count. Total followers are a point-in-time stock. New, lost and net followers are changes over a period.

Denominator: None for counts.

Net followers = New followers − Lost followers

Best comparison window: Save the follower total at consistent daily or weekly boundaries. Use the platform's selected period for new and net followers.

What it can diagnose: Current audience size and whether the account gained or lost followers during a defined window.

What it cannot prove: Followers are not the same as viewers. A follower may not see a post, while many viewers may not follow the account. An ending follower count also cannot reconstruct the exact follower base that existed when an older post was published.

Follower growth and follower conversion

Follower growth turns two follower snapshots into a change measure.

Absolute follower growth = Ending followers − Starting followers

Follower growth rate (%) = (Ending followers − Starting followers) ÷ Starting followers × 100

Define every variable. Starting followers is the account's follower count at the beginning of the window. Ending followers is the count at the end. If starting followers are zero, the percentage formula is undefined. Report the absolute change instead.

Native or calculated: Net followers may be native. Follower growth rate is normally calculated.

Best comparison window: Use equal periods such as week over week or one 28-day window against the preceding 28 days. Keep account and campaign conditions in the notes.

What it can diagnose: The pace of net audience change relative to the starting base.

What it cannot prove: Growth does not identify which posts, profile visits or outside events caused the change.

A post-level follow rate is another calculated diagnostic:

Post follow rate by views (%) = New followers attributed to the post ÷ Post views × 100

This is not a true unique-person conversion rate because the denominator contains views. If the same surface supplies post-attributed new followers and unique viewers, you may use unique viewers as the denominator, but label the result clearly. Do not publish a universal "good" follower conversion rate without a documented dataset, cohort, observation window and method.

Engagement rate

Engagement rate is not one universal TikTok metric. It is a family of formulas that divides interactions by a declared exposure or audience base.

TikTok One provides a native view-based engagement rate for supported projects:

Engagement rate by views (%) = (Likes + Comments + Shares) ÷ Video views × 100

In this formula, likes, comments and shares are interaction counts from the same scope. Video views is the view count for that scope. TikTok One's documented formula does not include Favorites. If you add Favorites, call it an expanded engagement rate and state the numerator.

A common creator-calculated alternative is follower-based:

Engagement rate by followers (%) = (Likes + Comments + Shares) ÷ Followers × 100

For a single post, define whether Followers is the count at publication or a fixed reporting snapshot. For a group of posts, do not mix summed interactions with one follower snapshot without explaining the method.

The two rates answer different questions. The view-based formula normalizes interactions by recorded exposure. The follower-based formula compares interactions with the size of the account's follower base, even though non-followers may generate many of those interactions. The detailed choice belongs in a separate comparison guide.

Tiksta's TikTok Engagement Rate Calculator provides a practical account-level calculation. This dictionary explains the broader definitions and denominator choices around that result.

Best comparison window: Use the same formula, interaction set, source, post age and content cohort.

What it can diagnose: Whether an apparent interaction change remains after normalizing by views or followers.

What it cannot prove: Engagement rate does not reveal a hidden distribution score. It can rise because interactions increased, because the denominator fell or because both changed. There is no defensible universal "good engagement rate" without a relevant benchmark dataset and method.

How TikTok Metrics Relate to Each Other

The metrics form a measurement chain, not a guaranteed causal funnel.

Read the relationships from left to right only as reporting logic:

Unique viewers or reach → views → watch-time observations → interaction counts → follower outcomes

The arrows do not mean every reached viewer creates a view, every view creates an interaction or every interaction creates a follower in a trackable sequence. Different people may perform different actions. The fields may also come from different aggregation systems.

Some relationships are mathematical. Average watch time uses views as its denominator. Completion rate uses complete views over all views. View-based engagement rate uses interactions over views. Others are only analytical comparisons. A rise in shares beside a rise in reach is an association unless you have a design that can establish cause.

TikTok publicly describes viewing and interaction behavior as recommendation inputs, but it does not publish a fixed score that assigns a universal weight to each creator metric. Its current recommendation-system explanation also makes clear that recommendations depend on multiple factors and vary by user. Use analytics to describe observed behavior, not to reverse-engineer a formula TikTok has not disclosed.

Worked Example: One Video, Several Denominators

The following numbers are illustrative. They are not Tiksta research, a TikTok benchmark or a claim about typical performance.

Assume a 30-second video has the following values seven days after publication, all taken from the same reporting surface and window:

Average watch time

180,000 seconds ÷ 12,000 views = 15 seconds per view

The average recorded view contributed 15 seconds of watch time. That is half of the video's duration, but it does not mean every viewer watched exactly half.

Completion rate

3,600 complete views ÷ 12,000 views × 100 = 30%

Thirty percent of recorded views reached the end under this reporting definition. It is not automatically 30% of unique people.

Views per unique viewer

12,000 views ÷ 9,000 unique viewers = 1.33 views per unique viewer

This ratio shows that total views exceeded the deduplicated viewer count. It does not tell you which viewers replayed or how TikTok handled every session.

View-based engagement rate using TikTok One's documented interaction set

(540 likes + 60 comments + 120 shares) ÷ 12,000 views × 100 = 6%

Favorites are not included because the declared formula uses likes, comments and shares. An analyst who adds 180 Favorites would produce a different metric and must give it a different label.

Post follow rate by unique viewers

90 post-attributed new followers ÷ 9,000 unique viewers × 100 = 1%

This ratio is usable only because the example assumes post-attributed followers and unique viewers come from the same surface and window. It still does not establish why each person followed. No result above is "good" or "bad" until it is compared with a relevant cohort.

How We Measure TikTok Performance

A reliable measurement method makes every comparison reproducible. It also prevents a changed denominator from looking like changed content performance.

Select one data source

Choose TikTok Studio, Web Business Suite, TikTok One or an authorized API as the primary source for a report. Export the exact field names when possible. Do not fill a missing Studio field with an Ads Manager value unless the report is explicitly designed to combine paid and organic data.

Also record the retrieval time. TikTok One says some report areas update in real time while others update daily or every seven days and can carry a 24 to 48-hour delay. That means two exports taken at different times may not be final even when they cover the same dates.

Fix the observation window

Choose the period before looking at the result. For post comparisons, use a fixed age such as the first 24 hours or first 7 days after publication. For account trends, use equal calendar periods such as consecutive 28-day windows.

Do not compare a post's lifetime views with another post's first-week views. Do not divide lifetime interactions by followers measured only today and present the result as historical.

Record every denominator

Write the denominator beside every rate. A report should say engagement rate by views rather than only engagement rate. It should say completion rate based on video views rather than leaving readers to assume unique viewers.

If the denominator changes, the rate can change even when the numerator does not. Raw counts and rates should therefore appear together.

Separate post-level and account-level metrics

Post views, post Favorites and average watch time belong to individual content. Total followers and net follower growth belong to the account. You may compare the levels, but do not merge them into one total.

Aggregated unique metrics need extra care. Adding the unique viewers of ten posts double-counts anyone who watched more than one post. Use the account-level unique field when the question concerns the account.

Compare similar content cohorts

A cohort is a group of observations that share relevant characteristics. For TikTok content, useful cohort factors include:

• Content format, such as tutorial, commentary or short demonstration

• Topic

• Video length

• Publication period

• Account size at publication

• Traffic source, including whether paid distribution was involved

Keep the cohort definition narrow enough to make the comparison meaningful. A 12-second product clip and a 3-minute explanation can both be successful, but their completion rates do not measure the same viewing challenge. A post published during a major external event may not belong in the account's normal baseline.

Use the median when one post distorts the average

The arithmetic mean gives every value weight, so one viral post can pull an account average far away from the typical post. The median is the middle ranked value and is less affected by extreme observations. The NIST guidance on measures of location explains why the median can better represent the center of data with extreme tails.

Suppose five comparable posts receive 800, 900, 1,000, 1,100 and 100,000 views. Their mean is 20,760 views, while their median is 1,000. The mean describes total scale spread across five posts. The median gives a much better picture of the typical post in that small cohort. Report both when scale and typical performance matter.

Log factors that can distort the result

Keep notes for paid promotion, Spark Ads, deleted posts, privacy changes, moderation, outages, collaborations, major trend participation and external traffic. TikTok states that Spark Ads can attribute paid views, likes, comments, shares and follows to the original organic post. TikTok's Spark Ads documentation makes that a material source distinction, not a minor footnote.

Step-by-step measurement workflow

1. Define the question. Write one testable question, such as "Did tutorial videos hold attention longer than commentary videos during their first seven days?"

2. Choose the unit of analysis. Decide whether each row represents a post, day, account or project.

3. Select the source and exact fields. Record the platform surface, export time and field labels.

4. Fix the observation window. Use the same post age or calendar duration for every observation.

5. Declare the denominator. Write the full formula for every calculated rate.

6. Build the cohort. Match format, topic, length, publication period, account size and traffic source as closely as the question requires.

7. Check the data. Flag missing fields, delayed reporting, promotions, deletions and outliers. Do not silently remove unusual posts.

8. Compare counts, rates and distribution. Review raw totals beside rates. Use the median for typical performance and the mean when aggregate scale is relevant.

9. Turn the result into one action. Change one measurable element for the next comparable cohort, then repeat the same observation method.

Which TikTok Metrics Should You Check?

Each action is a measurement decision, not proof that the chosen change will cause the result. The next cohort provides the evidence.

Limitations and Common Misinterpretations

Mismatched time windows

TikTok numbers accumulate at different speeds and some public counts are lifetime snapshots. A seven-day rate and a lifetime count do not belong in the same comparison. Set the window before calculating.

Mixing account-level and video-level data

Account post views describe all included content. A video's average watch time describes one post. Account followers describe a profile snapshot. Keep each level separate and link them only through a stated method.

Comparing different content cohorts

Topic, format, duration, publication period, account size and traffic source can all change the measurement context. A benchmark is useful only when its comparison group resembles the content being assessed.

Relying on one viral outlier

One high-view post can dominate the mean, total watch time and account-level engagement totals. Show the median and distribution across posts before describing typical performance.

Treating totals as rates

Five hundred shares is a total. Five hundred shares divided by 10,000 views is a share rate. The total answers how many. The rate answers how many relative to a base. Neither replaces the other.

Treating correlation as causation

If videos with more shares also received more reach, the two metrics moved together. That does not prove shares caused the extra reach. Recommendation, audience fit, traffic source, topic and timing can move at the same time. TikTok has not published a fixed causal weight for likes, comments, shares, Favorites, watch time or completion.

Assuming every account sees the same fields

Analytics fields vary by account type, region, device, surface and product rollout. A metric can also appear under a different label. Record the exact label and definition from your interface. Do not present a calculated proxy as a native TikTok field.

Combining paid and organic activity without a note

Promote and Spark Ads can change the counts attached to an organic post. Separate organic, paid and combined reporting when the surface allows it. Otherwise, flag the promotion period and avoid comparing the post with an unpromoted cohort as if acquisition conditions were equal.

Assuming a dashboard value is final

Reporting delays, moderation, deleted interactions and later views can change a value after the first export. Use a consistent data-freeze rule, such as exporting each post after seven full days at the same local time.

Your Next Measurement Step

Create a sheet with the last 20 eligible posts and freeze every post at the same age. Record source, export time, video length, format, topic, traffic source, views, unique viewers, total watch time, average watch time, complete views or completion rate, likes, comments, shares, Favorites and post-attributed followers where available.

Group the posts into comparable cohorts. Calculate the median for views, average watch time, completion rate and the interaction rates you have explicitly defined. Choose one result that differs materially within a cohort, write one plausible explanation and change one measurable content element in the next matched batch. Use the same source, denominator and window when you review it.