Analytics only help when each number is attached to a decision. A metric you look at but never act on is entertainment. Here are the six that change what you make next, and the rule attached to each.
1. Impressions
How many times your thumbnail was shown. This is YouTube's vote of confidence, not your performance. Low impressions mean the system has not found an audience to test you on — usually a topic problem, not a packaging problem.
2. Click-through rate
The percentage of impressions that became views. Typical healthy ranges sit between four and ten percent, but comparison across channels is meaningless — compare only against your own median.
A high CTR on low impressions is the best possible signal: the packaging works and the video just needs more distribution, which usually arrives if retention holds.
Decision rule: below your median by a third or more, swap the thumbnail and rerun. Above your median with weak retention, your packaging is overpromising.
3. Average view duration and percentage
Duration in minutes matters for long-form; percentage matters for short videos. Track both, but judge a video against others of similar length, never against the channel average.
Decision rule: if percentage viewed drops as you make longer videos while absolute minutes rise, keep going — total watch time is what feeds recommendations.
4. The retention curve
The most actionable chart in Studio, and the least used. Read it in three places:
- The first thirty seconds — every video drops here. A fall steeper than about forty percent means the opening is not delivering the title's promise.
- Mid-video dips — scrub to the timestamp. Something specific lost them: a tangent, a slow explanation, an ad break in a bad place. Cut that pattern next time.
- Spikes — people rewatched or skipped forward to that moment. That is the content your audience actually wanted. Make more of it.
Compare curves, not numbers
Open three videos side by side and overlay their shapes mentally. Patterns repeat across uploads far more reliably than single-video numbers.
5. Returning viewers vs new viewers
A channel that only ever gains new viewers has no audience — it has a series of lottery tickets. Returning viewers are the number that predicts whether you still have a channel in a year.
Decision rule: if returning viewers are flat while views grow, your content is attracting strangers who have no reason to come back. Add a recurring format, a series, or a consistent point of view.
6. RPM (not CPM)
CPM is what advertisers pay per thousand ad impressions. RPM is what you actually earn per thousand views after the split and after unmonetised views. Only RPM is your number.
RPM varies enormously by topic and viewer location — finance, software and business content earn many times what entertainment does for the same view count.
Decision rule: if you need revenue rather than reach, shift topics toward higher-RPM subjects before trying to double your views.
What to mostly ignore
- Likes and comments counts — nice, weakly predictive. Do not optimise for them directly.
- Real-time view counter — pure anxiety with no decision attached.
- Subscriber count — a lagging indicator. Subscribers gained per video is slightly more useful.
- Traffic source percentages — interesting, rarely actionable unless search is a deliberate strategy.
A twenty-minute monthly review
- List last month's videos with impressions, CTR, average duration and returning viewers.
- Mark the top two and bottom two on CTR; note what the thumbnails had in common.
- Open the retention curve for the bottom two; find the exact moment people left.
- Write one sentence describing what you will repeat and one describing what you will stop.
- Put both at the top of next month's content plan.
That is the whole discipline. Analytics are not a scoreboard — they are a list of editing instructions for the next video.