How to analyse a YouTube script before you film it
Almost every retention problem is decided before the camera is switched on. The tangent that loses people at four minutes, the payoff that arrives ninety seconds late, the section that repeats what the viewer already understood — all of it exists on the page first, where fixing it costs a paragraph instead of a reshoot.
Updated September 2026 · By the Retti team
Why the script is the cheapest place to fix retention
There is a rough hierarchy of how expensive a retention fix is, and it only runs one way:
- Script: rewrite a paragraph.
- Edit: cut, reorder, or paper over with b-roll — limited by what you actually filmed.
- Published: nothing. The video is what it is. You learn something for next time.
Most creators do their serious analysis at the third stage, where the information is best and the options are gone. Reading the retention curve of a finished video is genuinely valuable, but it is a lesson for the next upload, not a repair to this one.
What to look for, in order
1. Where the promise is made, and whether it is kept
The opening sets an expectation. Find the exact line where the video delivers on it. If that line is more than a third of the way in, you have a structural problem no amount of editing will disguise — viewers are waiting rather than watching.
2. Sections with no forward motion
Read each section and ask what changes by the end of it. If the answer is "the viewer knows a bit more about the same thing", that section is a plateau. Plateaus are where the long, sagging dips in a retention curve come from — not from anything dramatic going wrong, just from nothing in particular happening for a while.
3. Open loops, and whether they ever close
A question planted early holds attention only as long as the viewer remembers it. A loop opened in the hook and left untouched for five hundred words has quietly stopped working. Either reference it or drop it.
4. Repetition you cannot hear
Scripts repeat far more than their authors notice, because each pass is written in a different sitting. Two sections making the same point in different words is one of the most common causes of a mid-video drop, and it is close to invisible when you read your own work in order.
5. Jargon that arrives unexplained
Every niche has terms that feel basic to the person writing and opaque to a portion of the audience. Each one is a small moment of exclusion, and they compound.
What a script cannot tell you
Script analysis is a prediction, not a measurement, and it is worth being clear about the limits. Delivery can rescue a flat paragraph or sink a strong one. A visual can carry a stretch that reads as thin on the page. Pacing in the edit changes everything about how a section lands.
So a script read is best understood as removing the problems that no delivery can fix — structure, order, repetition, missing payoffs — rather than as a forecast of the curve. The things it catches are the things that would still be wrong however well you performed them.
Doing it in a loop
The most useful pattern is not one analysis, it is several. Read, fix the biggest structural problem, read again. Each pass surfaces what the previous fix exposed, and problems that were hidden behind a bigger one become visible once it is gone.
The trap in that loop is that a score on its own tells you very little between passes. A number that moved from 66 to 66 can hide three problems fixed and four introduced. What matters is which specific issues resolved, which are still there, and what the edit created — so track the individual problems rather than the total.
Where Retti fits
Retti's script analyzer reviews a finished script against the structure that holds attention: where each drop-off risk sits, what the hook is doing, which sections plateau, what repeats, and what to change before filming. It works on a script you have already written — you do not have to have drafted it in Retti.
Because creators run it in a loop, a re-analysis of a script it has seen before reports what changed since the last pass: which problems you resolved, which are still flagged, and which ones the rewrite introduced. For the opening specifically, hook analysis goes deeper on the first fifteen seconds. Once the video exists, Edit Review covers the cut and Retention Lab reads the real curve. First analysis is free.
Frequently asked questions
Can you predict YouTube retention from a script?
Partly, and it is worth being honest about which part. Structural problems — a payoff that arrives too late, a section with no forward motion, an open loop that never closes, repetition — are visible on the page and would still hurt however well the video was performed. Delivery, visuals and edit pacing are not visible, and they matter. Treat a script read as removing the fixable problems rather than forecasting a number.
How long should a YouTube script be for a ten-minute video?
Roughly 1,300 to 1,600 words for most conversational narration, at about 130 to 160 words a minute. Faster formats run higher, and anything with significant visual sequences or silence runs lower. The reliable method is to read it aloud at normal pace and time it, because written estimates are almost always short.
Should I analyse my script or just film it and see?
Film-and-see gives better information at a far worse moment. The retention curve of a published video is the clearest diagnostic available, but by then the only thing you can change is the next video. Script-stage analysis catches the structural problems while they still cost a paragraph, which is why most creators end up doing both.
What causes mid-video drop-offs in a script?
Most often a section with no forward motion — the viewer is not confused or annoyed, just no longer gaining anything. The other frequent causes are repetition the writer cannot hear because each pass was written separately, a tangent that breaks the through-line, and a payoff that arrives after the audience stopped waiting for it.
Does analysing a script make it sound formulaic?
It can, if you let a framework overwrite your voice. The useful version flags where attention is likely to slip and leaves how to fix it to you, because the fix that suits a documentary is not the one that suits a challenge video. Be wary of any rewrite that makes your script sound like everyone else's.
How often should I re-analyse a script while writing?
After each substantial structural change rather than after every sentence. Problems hide behind bigger problems, so a pass immediately after fixing the largest issue usually surfaces something that was invisible before. What matters between passes is which specific issues cleared and which appeared, not whether an overall score moved.