Four numbers worth reading
YouTube Studio presents dozens of metrics and most channels look at one: subscribers. It’s the least useful number available. It’s lagging, it’s vanity, and it tells you nothing about what to do next.
These four do.
Click-through rate
What share of people who saw your thumbnail chose to watch. This is the first gate, nothing downstream matters if nobody enters.
New channels commonly sit around 2–3%. Anything sustained above 5–6% is strong. The variable is almost entirely thumbnail and title, and the fix is testing rather than theorising.
One caveat when reading it: a very high click-through rate on a video with poor retention usually means the thumbnail overpromised. The two numbers should be read together, never separately.
Audience retention
What share of the video people actually watched, and, more usefully, the graph showing exactly where they left.
The shape tells you the problem. A cliff in the first thirty seconds is an opening problem. A steady decline is a length or pacing problem. A drop at a specific timestamp is usually a specific thing: a digression, a slow section, an unfulfilled promise.
This is the most actionable graph on the platform because it points at the exact second something went wrong.
Average view duration and watch time
Retention as a percentage, converted into minutes. It matters because watch time is what monetisation thresholds are measured in, and because the platform’s distribution favours videos that hold people.
A short video with excellent retention and a long video with mediocre retention can produce the same watch time. Which is preferable depends on your goals, but you can’t judge either without looking at both numbers.
Traffic sources
Where views come from: search, suggested, browse, external, shorts feed.
This tells you what kind of channel you actually have, which is often not what you think. A channel getting most views from search is an evergreen library and should be optimising titles and descriptions. One getting views from suggested is riding recommendation and should be optimising for session time and consistency of topic. The strategies genuinely differ.
What to do with them monthly
Look at all four together, for every video, once a month. The pattern is what matters, not any single video’s performance.
Then change one thing. Channels that change five variables at once learn nothing, because nothing can be attributed. Change the thumbnail approach for a month, or the opening structure, or the length, one at a time, and the data will tell you whether it worked.
The number to stop looking at
Subscribers. It’s a consequence of everything above, it moves slowly, and watching it daily produces anxiety rather than decisions. Check it monthly, and only as context.
Reading these properly is what our analytics and optimisation work covers, and it’s the first thing we do on any engagement, including commission partnerships, where our fee depends on getting it right. Send us the channel and we’ll tell you which of the four is your problem.