Start with the metric and its source
Record the platform, the exact metric name, the source and the collection date. A view count is not automatically a count of unique people. Views, accounts reached and watch time answer different questions; preserve the platform’s definitions instead of relabelling everything as reach.
Creator-authorized analytics can contain information absent from a public profile. TikTok, for example, provides account and individual post analytics through TikTok Studio, with availability varying by experience, region and eligibility. A handle alone does not guarantee access to every metric. TikTok: tools for creators.
Show the mean alongside the median
For posts with a known view count, the mean is the sum of those counts divided by the number of those posts. Exclude posts with missing counts from both calculations. The median is the middle value after sorting the known counts; for an even number of values, average the two middle values. Extreme observations affect the mean more strongly. NIST: measures of location.
Consider a hypothetical five-video sample with 2,000, 3,000, 4,000, 5,000 and 86,000 views. Its mean is 20,000; its median is 4,000. Both are correct. The mean reflects all recorded views, while the median better represents the middle of this particular sample. Show both, with the five-post sample size, rather than presenting 20,000 as an expected result.
Separate publication dates from measurement dates
A publication window decides which posts qualify. An observation window decides when you measure each post. “Videos published in August, measured on 1 September” gives an early-August video much longer to accumulate views than a late-August one.
For a fairer operational comparison, you might measure each eligible video seven days after publication. Seven days is an example protocol, not a platform standard or a universal optimum. Choose a period that suits the campaign and repeat it consistently. If historical snapshots are unavailable, keep the current counts, disclose the varying post ages and avoid calling the comparison age-adjusted. Reconstructing a seven-day count from today’s total would invent data.
Define inclusion before seeing the winners
Write down whether the sample includes all eligible videos, the latest eligible videos, or a selected campaign. Record exclusions such as unavailable posts, duplicate records and unsupported formats. Keep missing values separate from genuine zeros. Label paid amplification and collaborations when that information is known; mark it unknown when it is not.
Do not silently remove a viral post because it changes the average. If it answers a separate question, show an additional clearly labelled comparison without it, alongside the complete sample. Likewise, pinned videos should not become a substitute for a dated sample: they may represent selected highlights rather than recent activity.
Compare like with like
Keep Instagram and TikTok summaries separate unless you have explicitly harmonized their definitions. Within a platform, inspect format, topic and distribution conditions before comparing creators. A sponsored tutorial and an unrelated organic comedy clip may have different jobs.
Include the median, mean, minimum, maximum, sample size and eligible date range in your working report. These are descriptive statistics, not a confidence interval or a forecast. Five observations can expose a pattern worth investigating; they do not establish a stable performance promise. Show the individual posts when the sample is small enough to inspect.
Turn the number into a better question
Return to the hypothetical 86,000-view video. Check its topic, opening, length and publication time, along with what you know about its distribution. Identify something worth testing again; the sample mean of 20,000 views is not a forecast.
A useful report can end with a specific test: “This demonstration outperformed the other videos in the small sample. We do not know whether it had paid support. Test another demonstration and compare the views recorded seven days after publication.”