Compare Episode Retention to Find What Holds Your Audience

Inside a series, your best retention map is the other episodes. A single episode's curve tells you whether one video worked; a stack of episode curves tells you which of your recurring section shapes actually carry people. That comparison is a creative asset nobody exports and everybody should.

I run a three-episode stack whenever a series finishes a cycle: graphs open, episode lengths normalized, x-axes matched by percentage of runtime. Divergences become obvious fast, and so do the patterns shared across all three.

Normalize before you compare

Episode lengths differ, so match the x-axes by percentage of total runtime instead of by timestamp. A shorter episode still shows you which structural third it loses people in, as long as the comparison stays relative.

At the forty, sixty, and eighty percent marks, name the section each episode passes at that point. That is when the structural pattern appears. If all three sag around the same physical part of the runtime, the sag belongs to the format, not to any single script.

What episode stacks expose

• A recurring cliff at the same runtime percentage points to a repeated section type you are keeping out of habit.
• One episode beating the others at a specific section gives you the template for what that section should become.
• An episode that opens worse and finishes better means your strongest material lives past the midpoint, so lead the following episode with that shape.
• A first episode walking down while later episodes recover signals that your series rewards continuity and the return viewers are the ones holding the line.

One episode is a weather report. The stack is the climate, and the climate is what your return viewers actually signed up for.

Turn the comparison into the next script

Use the winning episode's section layout as the beat sheet for the next entry. If episode three held people through its halfway story beat and the others failed, write episode five around the same shape. The stack turns a popularity contest into a spec you can write to.

Do this while the series is current, not after it has stalled. Feedback from the stack is the fastest reliable signal you have, faster than comments, because it reflects what happened rather than what someone thought about it.

Watch the arc rise

When a cycle closes, run the stack again and check whether the recurring dips moved upward over time. A series that learns episode over episode shows an improving arc in the curves, and that stacked evidence is what you show a team, a sponsor, or your own doubting self before the next run.

A single cliff makes you react. A stack of cliffs makes you build. Series retention is the only place the graph hands you your own archive of causes to write against.