Writing Shorts Scripts Around the Retention Curve

The retention curve is a graph of how many viewers are still watching at each second. Every drop is a decision someone made to leave, and most of those decisions trace back to a line in the script. The curve is feedback, not a grade. Treat it as a map of where the writing failed.

Read the curve before you touch the edit. The fix is usually a rewrite, not a transition. Editing can hide a bad line, but it cannot make the line worth hearing.

I keep the raw numbers in a spreadsheet so I can compare across weeks. Memory flatters a good clip and punishes a bad one, and the sheet does neither.

The three shapes you will see

A cliff in the first two seconds means the hook did not land. A slow slide through the middle means a beat ran too long or turned dry. A spike at the end means the loop is pulling people back in. Three shapes point to three different problems.

Each shape has a different cause, so each one has a different fix. Treating every dip the same way is why some edits never move the number no matter how many cuts get added. Diagnosis comes before treatment.

I label the shape before I decide anything. Naming the problem keeps me from rewriting a hook when the middle is what actually failed. Most wrong fixes are just right fixes aimed at the wrong second.

The curve is not judging your editing. It is pointing at your sentences.

Map the script to the timeline

I time each line and mark it on the curve. When a dip lines up with a specific sentence, I know exactly which sentence to rewrite. Precision feels slow and saves hours.

This turns a vague sense that the middle drags into a specific target. The middle does not drag, one paragraph inside it does, and that paragraph has an address. You can rewrite an address. You cannot rewrite a feeling.

Common causes of a mid-clip slide

• A setup line that repeats something the viewer already understood.
• A tangent that is interesting to you and irrelevant to the promise.
• A pause long enough to feel like the clip ended early.
• A change of topic without a signal that a change is coming.

Write for the second watch

A rise at the very end usually means people are rewatching, often because the ending loops cleanly. If the curve falls instead, the last line closed the clip and gave permission to leave. Endings decide whether a view becomes two.

I check the final two seconds of every clip. If the number slopes up, the ending earned its place. If it craters, the ending is doing the opposite of a loop. The last line is a lever.

One curve is a data point, not a verdict, so I wait for the same dip across three clips before I change a habit. Patterns survive noise. Single readings do not.

Key takeaways

• Name the shape of the curve before you edit.
• Line up each dip with the sentence that caused it.
• Cut setup lines, tangents, and dead pauses first.
• Give the clip a loop so the curve rises at the end.