Reading one video to brief the next
A retention graph is a transcript of your edit; a few recurring shapes in it each point to a different question about the video you have not filmed yet.
You have looked at the graph. It goes down. You felt slightly worse and closed the tab, which is what most people do, because a line that falls looks like a verdict and nobody enjoys a verdict about their own afternoon.
It is not a verdict. A retention graph trends downwards, because most people leave before the end and most arrivals are at the start, though chapter and timestamped links do drop some viewers into the middle. That is a tendency and not a law, which is why the places the line goes back up are worth more attention than the slope itself. The information is not in the fall but in its shape, and a handful of recurring shapes are worth learning to read, though a real graph is usually a mix of them and no shape maps cleanly to a single cause.
By the end of this you should be able to open one video, read it in five minutes, and come out with three things to try on the next one.
What is a retention graph actually showing you?
The proportion of people still watching at each moment, from the start of the video to the end.
That is all. It is not a quality score, it does not compare you with anybody else, and the number at the end means little on its own, because a twenty minute video and a three minute one cannot be judged on the same scale.
What it contains is a record of every decision you made in the edit, in the order you made them. Read that way it stops being a mark and becomes a transcript, and a transcript can be acted on.
Why is a cliff in the first seconds not a content problem?
Because those people never got as far as the content.
A sharp drop at the very start usually means people opened the video and left almost at once. They did not dislike your explanation, because they never heard it. They discovered the video was not the thing they thought they had clicked, and went back.
That points to packaging, and packaging means the title and the thumbnail, the two things that made a promise before anybody pressed play. Either the promise was for a different video, or it was vague enough that people opened it to find out.
The fix is rarely in the edit. It is usually in the pair of things that got them there, and sometimes in the first sentence, which should confirm the promise rather than restate your channel name.
What does a steady slide mean?
Pacing, most often, and the most common shape by a distance.
A line that falls evenly with no particular event in it is usually a video where people are gradually deciding they have got what they came for. Nothing dramatic went wrong. It stopped repaying attention at roughly the same rate all the way down.
The instruction is unglamorous and it works: find the material that repeats. A steady slide is often a video where the same point is made in three places, or where each section takes longer to arrive at its idea than the one before. Tightening does more for that line than any change in style.
Is a flat stretch the good news?
Yes, and almost nobody looks for it.
Everybody stares at the drops. The flat sections are where people stayed all the way through something, which means that stretch was doing what it was meant to. Find it and watch it back.
Then ask what was happening. Was it the part where you finally showed the thing rather than describing it? Where somebody could see hands doing the work? Where you named a mistake out loud? Whatever it was, that is the material to make more of, and a far more reliable instruction than anything a drop gives you.
What is a bump, and which kind is it?
A rise in the line means people went back and watched something twice, for one of two reasons. Either the section was unclear and they had to see it again to understand it. Or it was worth seeing twice, because something happened quickly, or looked good, or was surprising.
Telling them apart requires no data. Watch the section. If it contains an instruction delivered once, at speed, with the important part happening off the edge of the frame, it was probably unclear, and the fix is to slow it down or show it again. If it contains a result, a reveal or a moment of movement, it was more likely rewatched on purpose, and the fix is to give it more room and put more such moments in future videos.
Two identical shapes, two opposite instructions. This is why the graph on its own is not an answer, and why every reading here is a question to check against the footage rather than a verdict to act on blind.
How do you turn all this into three decisions?
Open one video, take the four readings above, and write three lines. Not observations. Things to try, addressed to whoever films the next one, including you.
One about packaging, from the opening seconds. One about structure, from the slide and the flat stretch. One about a single moment, from the bump. Three is the most you will actually apply, and a list of eleven improvements produces the video you made last time.
Then check it against something the graph does not know. Read the comments on the same video and notice whether anybody mentions the sound. People rarely praise good audio and reliably complain about bad, so a comment about the sound is worth more than a compliment about the edit.
When this studio prepares a brief for a partner's next video, that is its shape: which sections were rewatched, what the invitation at the end produced, and whether the comments mention the sound. Short on purpose, because a brief nobody reads changes nothing.
What goes into the next brief
Three lines and a note about the audio, written before you forget which video they came from.
Which leaves one decision, and it is about frequency rather than analysis. Doing this once, after a video that disappointed you, teaches you about that video. Doing it after every video for three months teaches you about your channel, because the same instruction turns up on three separate sheets of paper and stops being an opinion.
Keep reading
This is a service, and the method is written down.
Everything above came out of doing the work rather than writing about it. If you want the method instead of the story, it runs in order on one page.
Start with a free audit
Tell us where your content is now. We will come back with what we would change and what result to expect.
A person reads the channel and writes the audit by hand: a considered read typically takes three working days. That is the usual shape, not a promised turnaround. We use these details only to reply to you: no lists, no lurking.
What you will get
A fit snapshot: where your channel stands, and whether we are a match.
Two to three opportunities: specific, prioritised, yours to keep.
A recommended next step, even if that step is not us.
The audit is free and commits you to nothing: nobody follows up with a call you did not ask for.