Retention Curves Decoded: Where Viewers Drop and Why
Your retention graph is a trail of decisions. Learn to read the shape of it — and the script edits that flatten the drops.
A graph that holds steady and then drops like a stone at one timestamp is the best news in retention analytics. It means most of your video is working and one minute killed you. Find that minute, look at what moved on screen at that second, and you have a leak you can patch in an afternoon.
Whole-video diagnosis is for people with a hundred problems. A single cliff means you have exactly one.
Studio lets you hover over the curve to get second-by-second values. When you spot a steep drop, record the exact second and go to that frame in your export. Look at the shot, the audio, and the on-screen text together, because the cause can hide in any of them.
Do not look at the clip on your timeline; look at the delivered file. Compression, burned-in captions, and your actual color grade are all part of what the viewer saw. The delivered file preserves the honest experience.
When the line is flat and then falls, the viewer was with you until that exact moment. Reconstructing that moment is reconstructing what you lost.
Some single-minute cliffs are artifacts. A mid-roll ad insertion shows up as a dip depending on the hour, and a chapter card can register as a small step. Reopen the curve a week later with more data before you cut apart a healthy segment.
A dip that repeats at the same timestamp across all your uploads is the platform, not you. A dip that moves along with your edit is you. Move the segment to a different spot in a test upload and watch whether the cliff follows it.
Your retention graph is a trail of decisions. Learn to read the shape of it — and the script edits that flatten the drops.
Cuts per minute tell you almost nothing about pace. Beat density, the rate at which something changes for the viewer, is the number that actually decides whether a video drags.
Viewers leave at seams, not mid-sentence. Close one loop and open the next in the same breath, and the whole video stops feeling like it has exits.
Analyze a channel, extract its structural DNA, and generate scripts engineered from patterns that actually work.
Analyze a Channel