Most advice treats YouTube's engagement metrics as separate projects: one checklist for likes, another for comments, another for watch hours. They are not separate. They are five readings of the same thing, whether the video held someone long enough to make them act.
That matters because it changes the order you work in. Chasing likes on a video nobody finishes is wasted effort; the like is downstream of the retention. This covers what genuinely moves each metric, and where each one sits in that chain.
Watch time is upstream of everything else
Every other metric on this page depends on people still being there. A viewer who leaves at forty seconds does not like, comment, or share, so retention is not one metric among five, it is the constraint on the other four.
Open the retention graph in YouTube Studio and look for the cliff. Almost every video has one, and its timestamp tells you more than the average view duration does. A drop in the first fifteen seconds is a packaging problem: the title or thumbnail promised something the opening did not deliver. A drop halfway through is a pacing problem.
- Find the steepest drop in the retention graph and watch that exact moment back
- A first-15-seconds cliff is almost always a title/thumbnail mismatch, not a content problem
- Cut the recap, returning viewers already know, and new viewers have not committed yet
- Compare retention across your own videos, never against another channel's
Likes follow a moment worth rewarding
People like a video when it does something specific for them. Solves the problem, lands the joke, confirms what they suspected. The like is a reaction to a moment, which is why generic "smash that like button" openings underperform: they ask before anything has happened.
Ask after the payoff instead, once. If the useful part of your video is at 4:20, the ask belongs at about 4:30 and nowhere else. Repeating it does not double the likes; it costs you retention, which costs you everything downstream.
- Ask once, immediately after the video has actually delivered something
- Name the reason. "if that fixed it for you" beats "don't forget to like"
- Check the like-to-view ratio across your own catalogue to find which topics land
- A rising dislike ratio on one video usually means the thumbnail over-promised
Comments need a question only your viewer can answer
"Let me know what you think" gets nothing because it asks for everything. A comment happens when someone has a specific answer and a reason to type it, which means the question has to be narrow enough to answer in one line.
The most reliable version is a question about the viewer's own setup: what they use, what went wrong for them, which of two options they picked. Pinning your own comment with the question restated gives latecomers somewhere obvious to reply, and replying to the first several yourself is what makes the section look alive rather than abandoned.
- Ask something answerable in under ten words
- Questions about the viewer's own situation outperform questions about the video
- Pin the question so it survives the sort order
- Reply to early comments: an empty-looking thread stays empty
Shares are private, and that changes what earns them
A share almost always happens in a direct message, not in public. Somebody sends your video to one person because it is useful to that person specifically, so shares reward content with a clear recipient, not content with broad appeal.
That is why tutorials, comparisons and "here's how this actually works" videos out-share entertainment on the same channel. If you want more shares, make something one viewer would want a named friend to see, and make the value obvious from the first frame so the recipient opens it.
- Videos that answer a question someone else is asking get sent to that person
- Timestamps make a long video shareable. People send the moment, not the hour
- A title that states the outcome travels better in a DM than a clever one
- Check the sharing surface in Studio to see whether it is DMs or feeds
Watch hours: the monetisation threshold is a by-product
If you are working toward the Partner Programme threshold, the honest answer is that hours accumulate from the four things above rather than from anything you can do to the number directly. Longer videos do not reliably produce more watch time. A fifteen-minute video watched for two minutes loses to a six-minute video watched to the end.
The two levers that genuinely compound are a back catalogue that keeps earning and a reason to watch the next video. Playlists and end screens matter here more than upload frequency, because they turn one view into a session.
- Watch hours are the sum of retention across the catalogue, not a target to chase
- Playlists convert a single view into a session. The highest-leverage change most channels skip
- End screens work when they promise something specific, not "watch more"
- Padding a video to hit a length reduces retention and therefore watch hours
The order to work in
Because these metrics sit in a chain, fixing them out of order wastes effort. Retention first, because nothing else happens without it. Then packaging. Title and thumbnail. Because that decides how many people enter the video at all. Only then the asks: like, comment, share.
One caveat worth stating plainly: no tactic here overcomes a video nobody wanted. If retention is flat and low across everything you publish, the problem is the topic selection, and no amount of engagement prompting will fix it.
- 1. Retention: find the cliff, fix that moment
- 2. Packaging. Make the title and thumbnail match what the video delivers
- 3. The asks. One like prompt after the payoff, one narrow question for comments
- 4. Sessions: playlists and end screens to compound watch hours
Frequently asked questions
Read next: How to Revive a Dead YouTube Channel, or Start Fresh. For more, browse our YouTube guides or all posts by Lyra Wilson.
