Measuring Shorts Performance Without Fooling Yourself

Views tell you almost nothing about a Short. A clip can rack up 500k views and still be a failure if everyone leaves in two seconds. What you want is a small set of numbers read together, not one big number read alone.

I keep four on a dashboard and ignore the rest: average view duration, the retention curve, shares per thousand views, and profile visits per thousand views. Each one answers a different question.

Average view duration and the first second

Average view duration is the headline. If a 30-second clip holds people for 20 seconds, the content is doing its job. If it holds for six, the hook is the problem, not the body.

Read it with the retention curve. A cliff right at the start means the hook failed. A slow slide means the middle sagged. A rise near the end means the loop is working and people are watching twice.

One number is a rumor. Four numbers read together are a diagnosis.

Shares tell you if it is worth passing on

Shares per thousand views is the honesty metric for a Short. Views can come from the algorithm shoving a clip in front of people who scroll past. Shares only happen when someone decides a friend should see this.

A share rate I like to see is around 1 to 2 per thousand. Below that, the clip might be pleasant but forgettable. Well above it, you have something people use to say something about themselves.

The numbers that do not mean what they seem

I have watched channels chase views into a hole, publishing more and more of whatever got a spike, until the whole feed was content their real audience did not want.

• Likes are cheap on vertical feeds; a like costs less attention than a view.
• Comments spike on controversial clips that also tank retention. Do not confuse argument with interest.
• Subscriber gains from Shorts are real but small; watch profile visits instead.
• Raw views without duration are the most misleading number on the platform.

A weekly reading, not a daily one

Shorts numbers bounce hard. A clip that looks dead on day one can pick up on day nine when the feed decides to retest it. I review a batch weekly, compare it to the previous week, and only make a call after two weeks of data.

When a clip wins, I ask why and try to repeat the one thing that caused it. When it loses, I ask the same question. The dashboard is only useful if it changes the next script.