TikTok Content Performance: Compare Topics Fairly

This guide follows a fictional ceramics teacher. Use the Content Performance Analyzer for local comparisons, not native account lookup or automatic creative recommendations.

Start with a decision smaller than your whole account

A useful TikTok content performance review starts with a question you can answer from the available records. Instead of asking whether the whole account is doing well, ask which recurring subject deserves another small batch. The analyzer compares supplied posts by topic or format using one selected measure. It does not inspect videos, read captions, identify trends, or explain distribution. That narrower job is valuable: a table of individual results can show whether an exciting headline number represents several posts or just one unusual result. Decide what you need to learn before selecting the posts and looking at their counts.

Fictional scenario: Talia teaches ceramics and runs one small workshop account. She has twelve organic videos covering glaze troubleshooting, studio setup, and clay reclaiming. Four posts belong to each topic, and she recorded their counts at a declared age of 72 hours. Her question is whether the next teaching batch should include more studio setup lessons. She chooses views as the first objective because those counts are available consistently. This is an example dataset, not a TikTok benchmark or a recommendation for every workshop. Another creator might reasonably begin with shares, comments, or a different supported measure.

Write the intended decision in an ordinary working document before opening the results. Talia writes that she will choose one topic for further exploration, not cancel the others or promise more bookings. This keeps the comparison connected to an action that the evidence can support. Views are viewing events under the source definition, not a count of different prospective students. A strong views result also does not establish that viewers understood the lesson. Those distinctions stop a simple descriptive comparison from becoming an unsupported statement about teaching quality, audience intent, or the reasons a particular video traveled further.

Collect a comparable snapshot before entering the numbers

Gather the post counts from records you can access, such as the individual content analytics available to your account. Check which source screen and measurement each column represents instead of assuming that every TikTok account exposes identical fields. The platform selection in this analyzer supplies context only; it does not sign in, fetch a profile, or verify your source. Keep a separate collection note describing where the figures came from. If one source calls something views and another uses the same word for a different counting rule, a shared column name alone does not make those figures comparable.

Observation age needs particular care when older and newer posts appear together. Talia records every example post after the same intended 72 hours of availability rather than copying twelve lifetime totals on one afternoon. Entering 72 in the analyzer declares that shared age; it does not adjust, normalize, or confirm the counts. If an older record was actually collected after five days, repair the source record or prepare a different comparison. The tool has no publication timestamp field that can reconstruct a matched snapshot. Keep collection timestamps outside the project when the review needs that additional audit trail.

Keep the traffic declaration equally honest. Organic, paid, and mixed describe the dataset you supply; they are not per-post switches or verified classifications. If Talia promoted one lesson, she prepares separate comparable datasets rather than expecting the analyzer to subtract the paid exposure. When separation is impossible, mixed is the appropriate declaration and the limitation remains part of the interpretation. Also decide the row unit in advance: one published video in this example. Do not insert both a video total and its daily slices as separate posts, because that would give the same publication several votes in the group statistics.

Make topic labels describe the lesson rather than the result

Use topic for what the lesson addresses and format for how it is presented. Talia labels a video about arranging shelves as studio setup and its presentation as demonstration. Another studio setup post might use a spoken explanation. Calling both fields studio tips would hide that distinction and make a later format comparison less useful. Choose a short label dictionary before entering performance counts whenever possible. The analyzer groups the supplied labels; it does not watch the videos or discover their subjects. The content pillar analysis guide explains how to make those manual choices consistent.

Assign one primary topic and one format to each row. When a video covers both reclaiming and storage, decide which reader question is its main purpose and record the secondary subject elsewhere in your own notes. Duplicating the post under both topics would not create an independent observation, and repeated IDs are not accepted. Avoid outcome labels such as viral lessons or weak posts. They build the result into the grouping rule before the comparison even begins. Useful labels remain meaningful when the counts are hidden: glaze troubleshooting, studio setup, and clay reclaiming pass that simple test.

Quick entry is suitable for a small review, while the CSV template helps with a prepared table. Supply a stable post ID, topic, format, and at least one supported count. Blank count cells mean missing, whereas zero means a known zero. Enter whole numbers rather than abbreviated values such as 1.8K, decimal rates, or numbers containing thousands separators. The project accepts up to 120 rows, with at most twelve topics and twelve formats. Those limits encourage a focused review, but they do not make a collection representative. Check the chosen records against your source list before pressing Analyze dataset.

Read the median beside the exciting outlier

Talia supplies glaze troubleshooting view counts of 1200, 1800, 2200, and 14800. Their median is 2000: after sorting, the middle pair is 1800 and 2200, whose average is 2000. Their mean is 5000 because the total of 20000 is divided by four. Both summaries are correct, but they answer slightly different questions. The mean reflects the large result strongly, while the median describes the middle of this small set. Neither reveals why the last post reached 14800 views. Keep that post visible rather than deleting it solely because it complicates the story you hoped to tell.

The four studio setup posts have 2600, 2800, 3000, and 3200 views, giving both a median and mean of 2900. Clay reclaiming has 1600, 2000, 2400, and 2800, giving a median and mean of 2200. The analyzer sorts groups by their medians, so studio setup appears above the other two for this objective. Glaze troubleshooting still has the highest individual maximum and the highest group mean. This is not a contradiction. It tells Talia that the most exceptional supplied post and the topic with the highest middle result are different parts of the same dataset.

Across all twelve eligible posts, the dataset median is 2500 views. Studio setup therefore has an index of 116, calculated as 2900 divided by 2500 times 100. That means its group median is 16 percent above this dataset baseline, not that the content is 16 percent better or will gain 16 percent more views next time. Glaze troubleshooting has an index of 80, and clay reclaiming has 88. This baseline includes the group being reviewed. It is an internal reference for the selected records, not an industry average, a separate control group, or an audience quality score.

Change objectives without changing what a rate means

Views and shares can lead to different descriptive rankings, so choose the objective that matches the current question. A lesson that attracts many starts may not receive many sharing actions, and a smaller lesson may generate a useful discussion. Raw shares compare supplied share counts per post; they do not divide by views. The analyzer has no general likes per view objective and no combined engagement score. Do not improvise one by moving a view count into the reach field. If the denominator required for a supported rate is unavailable or unsuitable, use an appropriate raw count and describe its limitations.

In a separate hypothetical rate illustration, suppose a source genuinely provides comparable reach and share counts for two posts. One has 30 shares and reach of 1500, giving 2 sharing actions per 100 reach units. Another has 40 shares and reach of 4000, giving 1 per 100. The second has more raw shares but the lower rate. These are actions relative to a supplied denominator, not percentages of unique people who decided to share. A person may perform multiple actions, and source definitions matter. The analyzer does not verify that a field imported as reach has the meaning the reviewer intended.

Rates are calculated for each eligible post before group summaries are calculated. A missing numerator or missing denominator excludes that row from the selected rate; a zero denominator also excludes it. A known zero numerator with a positive denominator produces a valid zero rate. Check the eligible and missing counts again after changing objectives, because the comparison may now describe a different subset. Four posts listed under a topic do not necessarily mean four usable rate observations. If only one has the required fields, a tidy median still rests on one post and should be read with that limitation in plain view.

Separate a topic pattern from the way posts were made

The first comparison may reveal a useful question about presentation rather than a final answer about subjects. Suppose all four studio setup posts are demonstrations, while the glaze troubleshooting set contains three spoken explanations and one demonstration. Topic and format are tangled in that collection. Switching Group by to format gives another descriptive view of the supplied rows, but it does not isolate the effect of demonstrations. Talia should inspect which topics contribute to each format and whether their production circumstances differ. The analyzer has no adjustment for video length, opening line, editing effort, or the audience reached by each publication.

Use the topic and format filters to inspect a narrower question when the data support it. Filtering to demonstration before grouping by topic compares only the supplied demonstrations. That may improve the practical relevance of the review, while reducing the number of eligible posts. Filters apply before the statistics, so they also change the dataset baseline and any available indexes. Do not quote an index from the filtered review as if it still used all twelve posts. Save the filter context with the result. The format comparison guide develops this distinction without claiming a controlled experiment.

Choose the minimum post threshold as a review reminder, not as a statistical certificate. If Talia selects four, a group with three eligible posts remains visible but is marked below that threshold. Increasing the setting does not remove the rows or make the remaining groups proven winners. Reducing it does not create more evidence. One visible group can still be summarized, but there is no real between-group comparison to make. Read sample count, missing count, median, mean, and range together so the clean sorting order does not conceal sparse coverage or a collection dominated by one unusual publication.

Turn the review into one manageable teaching experiment

Talia decides to prepare another small set of studio setup and glaze troubleshooting demonstrations. She keeps the teaching purpose, production effort, traffic choice, and intended collection age as similar as practical. Her working question is whether the middle views result remains higher for studio setup in another comparable batch. This is a creator planning choice based on the review, not an automatic recommendation generated by the tool. She does not promise herself the same ranking, because new topics, audience circumstances, and distribution can differ. The earlier observation is useful precisely because it points to something specific that can be checked again.

Keep the next batch small enough to finish and broad enough that one post does not carry the whole subject. For example, Talia can plan three new lessons for each of the two topics, then record each after her declared 72 hours. Three is an operational choice here, not a universal sufficient sample or a significance threshold. She also keeps a separate production note identifying any collaboration, unusual event, or change in distribution. The analyzer has no custom circumstances field, so those notes belong outside its rows. Avoid squeezing them into topic labels that would fragment the comparison into tiny groups.

Judge the next review against the question written before production. If setup again has a higher median, Talia has another descriptive reason to explore it, not proof that the topic caused the difference. If the pattern reverses, inspect the individual records and collection conditions rather than declaring the first review wrong. Both sets can be accurate snapshots. Teaching usefulness and workshop bookings require their own evidence outside this count comparison. Once the next lessons are planned, broader account work can sit in a separate SMTasker workflow without treating the local analyzer as a publishing connection or an account growth guarantee.

Save the context as carefully as the headline number

Save a project JSON when you need to reopen the supplied rows and settings later. The browser does not autosave this project, and the platform choice does not create cloud storage. A CSV replacement imports the required row template, not a complete native TikTok export, and it retains the current context settings. Recheck account label, observation age, traffic, objective, grouping, and filters after replacing a dataset. Opening a saved project is different: the project carries the whole context. These distinctions matter when someone prepares several reviews in succession and an old account label could otherwise follow a new table unnoticed.

After analysis, copied results, review CSV, and PDF help communicate the comparison. The review CSV is not the input template and is not a lossless substitute for the saved project. Review outputs retain the raw supplied dataset, including records outside active filters; the PDF also attaches complete project and review data. Check those contents before sharing a report, especially if post IDs or account labels are private working references. A filtered chart should not be mistaken for a file containing only those visible posts. Local calculation avoids sending the supplied counts to an analytics integration, but downloaded files can still disclose them when shared.

Finish the handoff with one short human conclusion: studio setup had the highest median views in this supplied twelve-post review, and Talia plans another comparable batch. Include the declared 72-hour age and organic context, plus the reminder that neither was independently verified. Any edit to the project invalidates its current results, so analyze again before creating a fresh report. Previously downloaded files do not update themselves. Keep the source collection note, project, and approved review together in your own folder. Return to the free tools directory for separate planning tasks instead of asking this comparison to answer questions outside its supplied counts.

Questions about TikTok Content Performance

Continue with another performance guide

Apply the comparison in the Content Performance & Topic Analyzer, or choose another guide below. Discover more resources in all free tools.