A post with 100,000 views could be an unusual result for one creator and an ordinary result for another. Without a comparison, the number tells us how large the recorded count is. It does not tell us how surprising the performance was.

A baseline supplies that comparison. YouTube already uses the idea in its analytics: overview reports can compare performance with what is typical for a channel. The broader question is useful across campaign reviews, but the comparison has to match the claim being made. YouTube: Channel performance overview

Choose the comparison before the verdict

Start with the question. Are we asking whether this post exceeded the creator’s recent results, whether a sponsored format held up against other sponsored work, or whether this creator performed differently from peers? Each requires a different set of reference posts.

For a creator’s own history, our suggested starting point is recent work on the same platform, in a comparable format, with the same observation window. Separate sponsored posts when that status can be established. Record paid promotion where it is known, and mark it unknown where it is not.

Comparing a three-day-old video with a post that has accumulated views for six months is a weak basis for judging the creative. So is mixing short clips and long tutorials without explaining their different jobs. YouTube’s own content guidance recommends comparing videos of the same type because audience behavior differs across formats. YouTube: Content analytics tips

The ideal comparison may not exist. Say which compromise was necessary and narrow the conclusion accordingly.

Keep the arithmetic inspectable

Here is an illustrative example, using invented numbers. A creator’s five comparable posts have recorded 20,000, 25,000, 30,000, 35,000 and 190,000 views at the chosen age. Their median is 30,000; their mean is 60,000. A campaign post with 45,000 views is 1.5 times the median but only 0.75 times the mean.

Neither calculation changes the underlying posts. The unusually large fifth result changes the mean substantially. Choosing one summary after seeing the campaign result could therefore change the verdict without adding evidence.

We would show the sample count, the distribution and the selected summary. A median can describe a typical result in a skewed sample, but it is not automatically the right target for every decision. If the business is buying total reach, occasional large outcomes may be part of the proposition it wants to understand.

For a simple ratio, divide the campaign metric by the chosen baseline metric. A zero denominator makes that ratio undefined; a very small denominator can make it unstable. Keep the absolute values alongside the ratio so a dramatic multiplier does not hide a small quantity.

A historical comparison is not a control group

A creator’s audience, topic mix and distribution can change between the reference period and the campaign. A promotion, event or platform change may affect both the sponsored post and the surrounding posts. Matching several observable features does not remove every difference.

Describe the result as performance relative to the selected history. Reserve causal language for a design that can support it. For example, Google’s Conversion Lift documentation describes an experiment comparing groups that see advertising with groups that do not. That is a different procedure from selecting earlier posts. Google: About Conversion Lift

Historical context can still improve a decision. It helps a reviewer identify which results need closer inspection and keeps account size from doing all the explanatory work.

Return to the actual content

After finding an unusual result, inspect what changed. Was the product demonstrated differently? Did the creator choose another format? Was distribution expanded? Did the comments concern the product or an unrelated event?

Treat these as competing explanations. If several posts with the same creative feature perform differently, retain that inconvenient evidence. A useful baseline should make it easier to challenge an attractive story, including the team’s own.

The next brief can then state a bounded proposition: repeat the demonstration format with a comparable creator group, preserve the observation window and examine whether the same pattern appears. That is a research plan, not a promised uplift.

Sources

All numerical examples are illustrative. This article proposes a review method; it does not report measured campaign results. Sources reviewed November 9, 2026.

Written by Brandnotomy Editorial.
Last reviewed Nov 09, 2026.

BaselinesMeasurementCreator performance