Every platform hands you more numbers than you can use, and the ones displayed most prominently are the least actionable. That is not malice; big numbers are satisfying and satisfying keeps people posting.
For actually deciding what to make next, three numbers matter and the rest are context.
The three that matter
| Number | What it means | What to do about it |
|---|---|---|
| Saves | Someone intends to come back to this | Make more of that exact thing |
| Sends | Someone thought of a specific person | This is your most shareable format |
| Watch-through | The hook and pacing worked | Fixable directly — see below |
| Comments | Enough investment to type | Useful, and reply to all of them |
| Views | Mostly distribution luck | Context only |
| Likes | Cheapest possible signal | Nearly ignorable |
Retention is the one you can act on directly
Compare against yourself
The most common analytics mistake is comparing a video to someone else's account, or to your own single best video. Both produce despair rather than information.
Use your own median instead. A video that did 1.4 times your median is worth studying; a video that did 0.6 is worth studying more. The absolute number is meaningless across accounts of different sizes and ages.
- Export or note the last 30 videos with saves, sends and watch-through.
- Find the median of each. Not the average — one viral video ruins an average.
- Mark every video that beat the median on saves. Those are your winners.
- Write down what the winners share. Format, topic, length, who they addressed.
- Repeat monthly. The pattern moves as the audience grows.
Read 60 days, not 7
A week of data is one or two videos and mostly luck. Distribution on short video is lumpy by design — the same video posted twice can perform very differently — and a weekly chart mostly measures that lumpiness.
Sixty days is enough to see a pattern and short enough that the pattern is still current. Anything longer and you are studying an audience you no longer have.
Numbers that mislead
- Follower count. It lags everything else and cannot be acted on.
- A single viral video. It tells you about that video, not about your account, and building on it usually fails.
- Impressions. Being shown is not being watched.
- Average watch time in seconds. A longer video has a longer average and may be worse. Use percentage.
- Engagement rate on a tiny sample. Below a few hundred views the number swings wildly and means nothing.
The short version
- Saves, sends and watch-through. Everything else is context.
- Look at the three-second cliff first — that is where the loss is.
- Compare against your own median, never another account.
- Use the median, not the average. One viral video ruins an average.
- Read 60 days. A week is one video's luck.
- Percentage watched, not seconds. A longer video flatters itself.