Views are the number everyone reports and the number that tells you least.
Between January and July this year, the videos we made for Strapi picked up 57,000 views. That's the number that goes in a deck. The number I'd actually defend is the other one: 459 hours of developers watching.
a view is a scroll, an hour is a decision
YouTube counts a view after about 30 seconds. Thirty seconds is somebody deciding your thumbnail was interesting and then leaving. It's a real signal about your packaging, and almost no signal about your product.
An hour is different. An hour is somebody with your docs open in the next tab, building along. Nobody spends an hour on a devtool video by accident. That's the closest thing to intent you can measure before a signup.
Put it the other way round: 57,000 views could be 57,000 people who bounced. 459 hours cannot be. The hours only exist if people stayed.
the numbers worth tracking
Watch time per video, not per channel. Channel-level totals hide the shape. What you want is which single video holds people for eight minutes, because that video is telling you which problem your audience actually has.
Then average view duration as a percentage. A four-minute video held to 60% is doing better work than a twenty-minute video held to 15%, even though the second one has more raw watch time. The percentage tells you whether the video was the right length for what it covered.
Then the retention graph, which has a cliff in it somewhere. That cliff is usually the moment the tutorial skipped a step, or the moment the presenter said "and obviously you'd already have X installed." Fix that spot and you've fixed the video.
Last, traffic to your docs from video. Most teams never wire this up. Put a tracked link in the description and you can stop arguing about whether video does anything at all.
what not to track
Subscriber count. It moves slowly, it's easy to inflate, and a subscriber who never watches anything is worth less than an anonymous viewer who watched twelve minutes. In our audit of 100 funded devtools, the median channel had 9,400 subscribers, and plenty of those channels hadn't posted in a year. The number stayed. The audience didn't.
Likes, for the same reason. Developers don't like things. They watch, they build, and they leave.
If you report one number upward this quarter, report hours watched. It's the one that survives the question "so what?"