View count is the number every platform shows first and the one every creator quotes. It is also the least useful number available, because it mostly measures how many people already follow you.
Here are seven that tell you something you can act on, roughly in order of how much they matter.
1. The multiple against your own baseline
Your median views over the last 90 days, divided into this post's views. A post at 3× did something. A post at 1.1× did nothing, no matter how large the raw number looks.
This is the only metric that separates the post from the audience, which is why everything on this site leads with it. The full definition is in how we measure a breakout.
2. Watch time as a percentage of length
Not total watch time — that rewards long videos. The share of the video the average viewer actually saw.
A 30-second clip watched to 80% beats a 90-second clip watched to 30%, even though the second has more total watch time. Percentage tells you whether the video held; total tells you whether it was long.
3. The three-second drop
How many people are still there after three seconds. This is the single most diagnosable number you have, because it isolates one thing: the opening.
If this is low, nothing else in the video matters — most people never reached it. Fix the first three seconds before changing anything else. It is also the cheapest thing to fix, because the fix is usually moving a sentence you already wrote.
Views tell you how many people arrived. Retention tells you whether arriving was worth it.
4. Saves
A save is someone saying "I want this later." It is the strongest signal of genuine value in short-form, and it is weighted heavily by most feeds because it is hard to fake and rarely accidental.
Saves correlate with practical content: instructions, lists, references. If your saves are low but views are fine, you are making content people enjoy and forget.
5. Shares, and especially sends
A share to a story is mild. A direct send to one person is strong — it means the viewer thought of someone specific. Sends are the closest thing short-form has to a recommendation.
Content that gets sent usually does one of two things: it is useful enough to pass on, or it articulates something the sender wanted to say and could not.
6. Follows per thousand views
Raw follower growth is noise. Follows divided by views tells you the rate at which strangers decide you are worth keeping.
A post with 50,000 views and 40 follows converted badly. A post with 8,000 views and 120 follows found exactly the right people. The second one is the format to repeat, and raw view count would have told you the opposite.
7. Comments that are not one word
Comment count is gameable and mostly meaningless — emoji and tags inflate it. What matters is whether people wrote sentences.
Real comments mean the video provoked a thought rather than a reflex. They are also the cheapest research available: the questions people ask in your comments are your next five videos, already validated.
What to do with all this
Do not track seven metrics. Track two: the multiple to find which posts worked, and the three-second drop to diagnose why the others did not.
The rest are worth checking when something surprising happens — a post with poor views and excellent saves is telling you something that view count alone would have buried.
Questions people ask
What is the single most useful metric for short-form?
The multiple against your own baseline: this post's views divided by your 90-day median. It is the only metric that separates what the post did from what your audience size would have produced anyway.
Why is watch time percentage better than total watch time?
Total watch time rewards length. A 30-second clip watched to 80% held attention better than a 90-second clip watched to 30%, even though the second has more total minutes.
Are saves more important than likes?
Generally yes. A save means someone wants the content later, which is hard to fake and rarely accidental, so most feeds weight it more heavily than a like.
How many metrics should I actually track?
Two. The multiple to find which posts worked, and the three-second retention drop to diagnose why the others did not.