YouTube Analytics: The 8 Metrics That Actually Predict Growth

YouTube Studio will show you somewhere north of forty metrics. Roughly eight of them are worth your time, and the other thirty-odd are actively harmful, not because they are wrong but because they give you something to look at instead of something to do.

You do not have an analytics problem. You have a too-much-analytics problem. Here are the eight numbers I check, what each is genuinely telling you, and the review rhythm that turns them into better videos instead of anxiety.

Why most analytics advice fails#

Two reasons, and both are about interpretation rather than data.

No single metric diagnoses anything. Click-through rate on its own is meaningless. Retention on its own is meaningless. They only become informative in combination, because each one describes a different stage of the same funnel. A number in isolation invites you to fix the wrong thing.

Almost everyone checks too often. Refreshing Studio an hour after publishing tells you nothing except how your subscribers behave, and it trains you to react to noise. I have watched creators change a title twice in one afternoon and permanently lose the ability to attribute anything to anything.

Everything below assumes you are reading numbers together and at sensible intervals. If you want the wider framing of where these fit, the five-stage system puts them in order.

The eight#

1. Impressions#

What it is: how many times YouTube showed your thumbnail to someone on the platform.

What it is really telling you: what YouTube thinks of your topic, and whether it has an audience it can associate with you.

This is the first number I look at on every video, because it splits every problem into two families. Low impressions means the problem is upstream — the subject has less demand than you assumed, or your channel is not clearly enough about anything for the recommendation system to know who to show it to. High impressions with low views means the problem is your packaging, and no amount of re-editing will help.

Do not compare impressions to other channels. Compare a video to your own recent videos. A video with a tenth of your usual impressions is telling you something about that topic specifically.

2. Impressions click-through rate#

What it is: the percentage of those impressions that turned into a view.

What it is really telling you: what viewers think of your packaging — the title and thumbnail together.

Most channels sit between 2% and 10%. But the raw number matters less than three contextual facts that almost nobody accounts for:

It varies by traffic source. Search click-through rates run higher than browse, because the viewer was already looking for the thing. A video that is mostly search traffic is not comparable to one that is mostly home page.

It varies with how many of your viewers are subscribers. People who already know you click more, so small channels with loyal audiences often post flattering numbers that fall as they reach strangers.

It falls as a video succeeds. This is the one that catches people out. As a video is shown to wider and wider audiences, click-through rate drops, because those wider audiences are less predisposed to your subject. A declining rate on a video that is growing fast is a sign of success, not failure.

The reason to care about this number is that it is the only multiplier you can improve on a video you have already published. Rewriting a title and thumbnail on an existing video is the cheapest experiment available — the rules are in thumbnails that get clicked and 27 title formulas.

3. Average view duration, in minutes#

What it is: how long the typical viewer watched, expressed in time.

What it is really telling you: what viewers thought of the video.

If you track only one number, track this one, because it combines whether the video was worth clicking and whether it was worth staying for.

Crucially, use the minutes figure rather than the percentage. Percentage does not survive comparison across different video lengths — six minutes held on a twenty-minute video is a better result than four minutes held on a ten-minute one, even though the percentages say the opposite. Any time you see someone compare retention percentages across different lengths, they are comparing things that cannot be compared.

4. Average percentage viewed#

What it is: the same thing as a proportion of the video.

What it is really telling you: whether the video was the right length for what it contained.

This is the one legitimate use of the percentage figure: as a length diagnostic, not a quality one. A consistently low percentage across videos of similar length usually means you are padding. A very high percentage combined with a short duration often means you could have said more.

Read it alongside the retention curve rather than on its own — the four curve shapes and what each means are in audience retention on YouTube.

5. Returning viewers#

What it is: how many people who watched had also watched something of yours in the previous 30 days.

What it is really telling you: whether you are building an audience or renting attention.

This is the most under-rated number in Studio and the one I would look at second if I could only pick two. Views can be bought with packaging. Subscribers can accumulate from a single viral video and mean nothing. Returning viewers is the closest thing to a measure of whether people actually want more from you.

A channel with rising views and flat returning viewers is a channel that is going to plateau, and usually the creator has no idea it is coming. It is also the number that exposes the Shorts trap, which I cover in the Shorts strategy guide: huge view counts, no growth in the people who come back.

6. Traffic sources#

What it is: where views came from — browse, suggested, search, external, Shorts feed, notifications.

What it is really telling you: which surface the video won on, and therefore what a follow-up should look like.

This is the most actionable report in Studio and the most ignored. A video that won on search should be followed by more search-shaped content, because you have found a query with real demand. A video that won on suggested tells you which videos yours is being shown next to — go and look at them, and make something that sits naturally beside them too. A video that won on browse means your packaging can compete on the home page, which is the hardest surface there is.

The four surfaces work differently enough that treating them as one number throws away most of the information. The algorithm piece covers why they behave so differently.

7. Unique viewers, and views per viewer#

What it is: roughly how many distinct people watched, and how many videos the average one watched.

What it is really telling you: whether a video is a doorway or a dead end.

Views per viewer above one means people watched a video and then watched another. That is a session, and sessions are disproportionately valuable because YouTube cares about the whole visit rather than a single video.

If a video brings lots of unique viewers and none of them watch anything else, it is a dead end. Usually the fix is a clearer next step: an end screen pointing at the obvious follow-up rather than "my latest video", and a genuine verbal reason to watch it.

8. Subscribers gained, per video#

What it is: net subscribers attributable to a specific video.

What it is really telling you: which videos make people want more.

Look at this per video, never as a channel total. The channel total tells you nothing you can act on. The per-video number tells you which format, subject and length converts a stranger into someone who wants the next one — and that is the thing to make more of.

Be aware that subscriber counts lag. A video that converts well often keeps producing subscribers for weeks, which is another reason not to judge anything at 24 hours.

What to ignore#

Being clear about what to skip is as useful as knowing what to watch.

Real-time views. Interesting, actionable never. It measures your subscribers' notification habits.

Likes and like ratio. A vanity signal that correlates with things you are already measuring better. Unless your ratio is unusually poor, which would suggest a mismatch worth investigating, it tells you nothing new.

Comment count. Correlates with how much you asked for comments, not with how good the video was.

Total channel watch time. Fine as a milestone, useless as a diagnostic, because it aggregates away everything that would tell you what to do next.

Impressions on day one. Too noisy to interpret. The system is still deciding who to test you on.

Anyone else's benchmarks. Screenshots from other creators are from different niches, different audience sizes and different traffic mixes. Compare your videos to your videos.

The review rhythm#

Three checkpoints. Nothing in between.

At 48 hours — check click-through rate and the Intro retention figure only. This is your packaging checkpoint, and it exists because a packaging fix still helps at this stage. If click-through rate is well below your average, rewrite the title and thumbnail now. Do not touch anything else.

At 7 days — the real read. All eight numbers. Compare against your previous five videos. This is where you form a hypothesis about what happened.

At 28 days — the verdict. Traffic sources have settled and suggested traffic, which builds slowly, has had time to appear. This is the number you record.

Build the one spreadsheet that matters#

Everything above becomes dramatically more useful with a simple record, and almost nobody keeps one.

One row per video. Columns: date, title, topic, format, length, impressions, click-through rate, average view duration in minutes, returning viewers, primary traffic source, and — the important one — what I changed this time.

After fifteen rows you will be able to see patterns that are completely invisible video to video: that your list videos out-retain your essays, that your search topics have half the impressions and twice the click-through rate, that every video over eighteen minutes underperforms.

That last column is what makes the sheet worth keeping. Without a record of what you changed, you have a log. With it, you have an experiment.

If you would rather work with the data programmatically, YouTube exposes most of this through the YouTube Analytics API (opens in a new tab), and Studio itself can export any report to a spreadsheet. But a manual sheet is genuinely better at the start, because the act of typing the numbers in makes you look at them.

Two questions to ask of any number#

When you are looking at a metric and are not sure what to do about it, these two questions will usually resolve it.

"Compared to what?" Almost every number in Studio is meaningless in isolation and informative in comparison. A 4% click-through rate is bad if your average is 7% and excellent if your average is 2.5%. Sixty seconds of average view duration is dreadful on a twenty-minute video and outstanding on a Short. Before you react to any figure, find the comparison that makes it interpretable — and the right comparison is nearly always your own recent videos, not a benchmark from the internet.

"Which stage does this describe?" Every metric belongs to a stage of the funnel: getting shown, getting chosen, getting watched, getting people back. Working out which stage a number describes tells you immediately whether the fix is in your topic selection, your packaging, your video, or your format. This is the discipline that stops people re-editing a video whose actual problem was that nobody was ever going to search for it.

Get in the habit of answering both before you change anything, and you will spend far less time fixing things that were not broken.

The one thing to take away#

Numbers do not improve channels. Attribution does.

The value of tracking is not that you know your click-through rate. It is that six months from now you can say "the videos that work for me are X, and I know that because I have fifteen rows of evidence." That is the difference between a channel that gets luckier over time and one that gets better.

If you are not yet at the point of optimizing — if the videos simply are not getting seen — start with why your YouTube videos get no views, which uses these same numbers as a diagnostic rather than a tracker. And for the official documentation on any report mentioned here, Get started with YouTube Analytics (opens in a new tab) is the reference.