Analytics
How to Read YouTube Analytics: The Complete Guide
How to read YouTube analytics properly: which metrics diagnose what, how to compare videos correctly, and the six inputs that make up a channel health score.
September 8, 2026 · 9 min read · By Christian Koch
Reading YouTube analytics means asking each metric the one question it can answer. Impressions tell you whether YouTube showed the video. Click-through rate tells you whether people chose it. Average view duration tells you whether it held them. Returning viewers tells you whether it built anything. Reading them as a single performance number hides every diagnosis worth having.
Key takeaways
- Compare every video against your own channel median at the same age. Cross-channel and cross-age comparisons produce false conclusions.
- Views and subscribers are outcomes. They summarize a result and diagnose nothing.
- Click-through rate is meaningless as a channel average and highly informative split by traffic source.
- The retention graph's shape carries more information than the retention percentage.
The four questions analytics can answer
Every report in YouTube Studio serves one of four questions, and knowing which one you are asking prevents most misreadings.
| Question | Metric | Report |
|---|---|---|
| Did YouTube show it? | Impressions | Reach |
| Did people choose it? | Impressions click-through rate | Reach |
| Did it hold them? | Average view duration, audience retention | Engagement |
| Did they come back? | Returning viewers, unique viewers | Audience |
Read in that order and stop at the first failure. Metrics below a broken gate describe a sample too distorted to interpret, which is why the click-through rate on a video with four hundred impressions tells you nothing. The full diagnostic version of this sits in the four-gate diagnostic.
Impressions: did YouTube show it?
Impressions count the times your thumbnail was displayed on YouTube surfaces. YouTube registers one when the thumbnail is shown for more than one second with at least 50% of it visible on screen.
The exclusions matter more than most people realize. Impressions do not count on external websites, in embedded players, on the mobile web, on YouTube Kids, on YouTube Music, in background tabs, or on thumbnails less than half visible (YouTube Help). A video with heavy external traffic therefore shows an impressions figure describing a minority of its actual reach, and its click-through rate describes an even smaller slice.
What to read it against: your own channel median at the same video age. Impressions well below that median means YouTube tested the video with a small audience and stopped expanding, which points at subject selection.
Click-through rate: did people choose it?
Impressions click-through rate is the share of impressions that turned into views. It is only interpretable split by traffic source.
Search viewers arrive with intent already established, subscriber-feed viewers already chose your channel, and home-feed viewers have neither. Those three convert at very different rates, and a channel-wide average blends them into a figure that describes none of them. A channel whose distribution shifts toward the home feed will watch its average fall while views rise, which reads as a decline and is the shape of a channel scaling.
Where to find it: Studio → select a video → Reach tab → break out by traffic source. Benchmarks in what a good CTR looks like.
Watch time and average view duration
Average view duration is the average minutes watched among viewers who stayed to watch, calculated from engaged views and their corresponding watch time. Watch time is the total hours viewers spent on the video.
The useful comparison is average view duration against video length, expressed as a percentage, since eight minutes on a ten-minute video and eight minutes on a forty-minute video describe entirely different outcomes. That said, absolute minutes matter for the recommendation system, which is why a long video with mediocre percentage retention can still outperform a short one that holds nearly everyone.
A caution on comparing across lengths. Shortening videos raises percentage retention and frequently lowers total watch time. Track both, and treat a rising percentage alongside falling minutes as a warning.
Most channels have the data and no reading of it. The free Channel Scorecard runs these six inputs against your channel and returns a single number with the underlying diagnosis attached, which is the version of this most people actually need.
Audience retention: the shape matters more than the number
The retention graph shows how well different moments held attention. Four shapes carry almost all of the diagnostic value.
A cliff in the first thirty seconds. The opening does something the packaging did not prepare the viewer for. Most often the video takes ninety seconds to reach what the thumbnail promised. This is the most common finding and the fastest to fix.
A steady, gentle decline. A healthy curve. The video is holding attention proportionally to its length. Weak views alongside this shape means the problem lives upstream in topic or packaging.
A collapse at a consistent timestamp across several videos. Structural. Something in the standard format loses people at the same point every time, frequently a sponsor read placed before the payoff or a stretch of setup that could be cut.
A spike partway through. Viewers are rewatching or jumping to that moment. Worth watching to see what it is, since it is often the thing the whole video should have been about.
Read the curve before reading the percentage. A 42% retention on a gentle decline and a 42% retention on a first-minute cliff describe two completely different videos.
Traffic sources: how the video reached people
The traffic source report shows which surfaces delivered your views, and it tells you whether YouTube understood the video the way you intended.
YouTube describes its recommendation system as working from viewing behaviour, likes, subscriptions and satisfaction survey feedback, alongside channel reputation and quality (How YouTube Works). The traffic source report is the visible output of that routing.
The four to watch:
- Browse features. The home feed. High volume, low intent, the main engine of scale.
- Suggested videos. Mid-session recommendations. A strong suggested share indicates the video sits well alongside related content.
- YouTube search. Low volume, high intent. Disproportionately valuable for business channels.
- External. Off-platform traffic. Views count, impressions largely do not, which distorts the CTR on these videos.
A video built for search that pulled mostly browse traffic was read differently than you intended. Neither outcome is a failure by itself; the divergence between plan and result is the finding.
New versus returning viewers
Returning viewers is the closest thing YouTube gives you to a read on whether the channel is building something. It matters most for channels with a commercial purpose.
High views with weak returning viewers describes rented attention: the recommendation system found an audience, satisfied them once, and gave them no reason to associate the satisfaction with your channel. Modest views with a strong returning share describes a smaller audience that is actually yours, and for lead generation the second is worth considerably more.
Track the ratio over quarters. Month-to-month movement here is noisy, and the trend across two quarters is honest. We cover the commercial version of this question in whether your channel is working.
The Channel Health Score
Six inputs, tracked monthly, rolled into one number you can watch move. This is the framework behind our Scorecard, and it works as a manual exercise too.
| # | Input | What it reads | Direction |
|---|---|---|---|
| 1 | Median impressions per video, last 90 days | Whether YouTube is expanding your distribution | Up |
| 2 | Median CTR, browse traffic only | Whether packaging earns strangers | Up |
| 3 | Median average view duration as % of length | Whether the content keeps its promise | Up or stable |
| 4 | Returning viewers as % of total | Whether the audience is becoming yours | Up |
| 5 | Ratio of best video to median video | Whether you are still finding outliers | Up early, narrowing later |
| 6 | Share of views from your top three subject pillars | Whether the channel has a centre | Up |
Input 5 is the one most channels have never looked at, and it is the earliest signal a strategy is working. A widening gap between your best video and your median means the loop is finding something. A channel where every video performs identically has not varied enough to learn anything.
Record all six on the first of each month. Six months of that data is worth more than any single-month deep dive, and it takes fifteen minutes to maintain.
Metrics that summarize without diagnosing
Total views, subscriber count and likes report outcomes. They belong in a summary and produce no actionable diagnosis on their own.
Subscriber count in particular misleads business channels, because subscribers acquired through a video unrelated to the commercial offer dilute the very signal you are trying to build. A channel gaining 4,000 subscribers who never watch the next video has weakened its subscriber-feed performance while the headline number improved.
The other trap is comparing your figures to another channel's. Different audiences, different traffic mixes, different subject areas, different ages. Your own median is the only benchmark with all of those held constant, and it is the comparison we use in the five-layer channel audit.
FAQ
Which YouTube analytics metrics actually matter?
Impressions, impressions click-through rate by traffic source, average view duration, audience retention shape, traffic source mix, and returning viewers. Those six answer the questions that lead to a decision. Views, subscribers and likes summarize results and diagnose nothing on their own.
How do I compare two YouTube videos correctly?
Compare them at the same age, against your own channel median, on one metric at a time. A video published two years ago has had two years to accumulate search impressions, so raw view counts across different ages describe elapsed time more than performance. YouTube Studio's shaded typical-performance range does most of this work automatically.
What is a good audience retention percentage?
The shape matters more than the percentage. A gentle decline holding 35% on a twenty-minute video is a strong result, while 55% on a three-minute video with a first-minute cliff is a weak one. Compare percentage retention only across videos of similar length, and read the curve before the number.
Why do my impressions and views not match?
Impressions count thumbnail displays on YouTube surfaces only, requiring more than one second of display with at least half the thumbnail visible. Views arriving from external sites, embedded players and the mobile web are largely excluded from impressions, so a video with heavy off-platform traffic will show views well out of proportion to its impressions.
How often should I check YouTube analytics?
Deep reads monthly, with a quick check on new videos at three weeks. Daily checking produces noise and encourages reacting to a video's first 48 hours, which mostly measures your existing subscribers. The six-input health score works best recorded once a month and read across quarters.
Where to start
Record the six health score inputs for your channel this month, then again next month. The direction of movement across two readings tells you more than any single deep dive.
If you want that produced for you with the diagnosis attached, the free Channel Scorecard does it.