Short-form analytics have a specific problem that other channels do not: several of the headline numbers are not comparable with themselves. The same video honestly produces three different engagement rates depending on the denominator, and the most quoted metric of all moves for reasons that have nothing to do with whether the content got better.
Engagement rate is three metrics wearing one name
Divided by followers, by views or by reach, one post returns results differing by an order of magnitude. Follower-based rates are a leftover from feeds that only showed your posts to people who followed you, which is not how any short-form platform works. These surfaces distribute to non-followers by default, so a video that travelled makes your engagement rate look terrible against a follower base it never needed.
Views is the honest denominator here. More important than which one you pick is never comparing a number computed one way against a published benchmark computed another, which is what almost every benchmark quietly invites.
Completion rate moves when you change length
Cutting a video in half will usually raise its completion rate without making it better. An account optimising for that number alone drifts steadily towards content too short to say anything, and the metric will report the drift as improvement the entire way.
It becomes useful as a pair with watch time. High completion and low watch time means the video was short. High completion and high watch time means it genuinely held people, and that combination is the one worth chasing.
Watch time in seconds is not comparable across videos of different lengths. Divide by the length and you have retention, which is.
Saves and sends are the underreported ones
A save is a statement of intent to use something later, which is a far higher bar than a like and a much better predictor of whether a video was valuable. A send is better still, because each one is a distribution event rather than only a signal: the video reaches someone through a route the platform did not choose.
Both are weighted heavily by the ranking systems and both are left out of most engagement formulas, which is a large part of why those formulas mislead. If you track two things beyond retention, track these.
The two worth steering by
For content decisions, retention past the first three seconds. It has more signal at low volume than anything else available, and it responds fastest to changes you actually make, which is what makes it useful as a steering input rather than a scorecard.
For account decisions, follower conversion rate: new followers divided by profile visits. Views are traffic and follows are conversion, and the profile is the landing page in between. Almost nobody measures it, which is why enormous effort goes into the video that earns a visit and almost none into the screen where the decision gets made.
What to ignore, and for how long
- Follower count. It is a lagging indicator of a lagging indicator and the easiest number to inflate with content that attracts the wrong audience.
- Any single post's performance. Distribution varies enormously per post, so one result is noise on every platform at every size.
- Published benchmarks. They are computed with denominators you cannot see, against accounts nothing like yours.
- Everything, for the first month of a new account. There is not enough volume for any of it to mean anything yet.
The glossary defines each of these properly, including the pairs that get confused with each other, which is usually the reason someone went looking for the definition in the first place.