What is a video retention curve: the retention curve explained, second by second
A video retention curve is the graph of what percentage of viewers are still watching at each moment of a video, starting at 100 percent on the first frame and falling as people drop off. It is the only video report that tells you where you lost somebody rather than how many you lost.
Also called an audience retention graph, a drop-off curve, or an engagement graph, depending on whose analytics you are looking at. Absolute retention plots the raw percentage still present. Relative retention compares that curve against other videos of similar length.
What does a normal retention curve look like?
Three parts, in this order:
- A steep initial drop. Every curve falls fast in the opening seconds as people decide this is not what they wanted. This is normal and unavoidable, and the size of the drop is the diagnostic.
- A long, gentle slope. Once the early leavers are gone, the remaining audience declines slowly and fairly evenly. A straight, shallow line through the body of the video is a healthy sign.
- A small tail. A modest drop near the end as people leave once the substance is finished, and sometimes a bump at the final seconds where the conclusion or the offer sits.
YouTube Help reports the first of these as an intro figure: "what percentage of your audience still watched your video after the first 30 seconds". For short business video the equivalent checkpoint is the first five to ten seconds.
Try Venta Capture on your own process
Build one flow for your highest-volume case type. Free, no credit card.
What a cliff in the first seconds means
A near-vertical fall in the opening three to five seconds is almost never a production problem. It is a promise problem, and it usually has one of four causes:
- Mismatch. The thumbnail, subject line or headline promised one thing and the video opened with another. The viewer resolves that in about two seconds.
- No orientation. The video starts mid-thought with no indication of who is speaking or why. Name the vehicle, the customer or the job in the first sentence.
- A branded intro. Any logo animation in front of the content donates the most valuable seconds you have to nobody. Cut it.
- Autoplay traffic. If the player starts unprompted, a large share of that first drop is people who never chose to watch. That is a measurement artefact, not an audience verdict, and it is why autoplay and click-to-play curves must never be compared.
How to read spikes and dips in the middle
YouTube's documentation is blunt about what the shape means: spikes are "moments in your video that were rewatched or shared", and dips highlight "moments in your video that were either skipped or moments where viewers stopped watching your video completely".
Both readings need care. A spike can mean a section was compelling or that it was unclear enough to need a second pass, and only watching the segment tells you which. A gradual dip is ordinary attrition. A sharp one at a specific timestamp is a moment you can go and look at, which is what makes this chart worth more than every average in the report.
A worked example
A service advisor's two-minute inspection video holds 92 percent of viewers at five seconds, then slides evenly to 61 percent at the one-minute mark. At 1:12 it falls 19 points in four seconds.
At 1:12 the advisor stops filming the worn component and starts reading the price breakdown off a screen. The customers who left had already seen the problem and did not need the recital. Moving the price to a written line beside the video, and ending on the part itself, is a change you can only find on a curve. The average watch time for both versions is nearly identical.
How the retention curve differs from watch time
Watch time is a summary of this curve, specifically the area underneath it. Two videos with the same watch time can have completely different shapes: one losing people steadily, one holding almost everyone and then collapsing at a single point. Only one of those has an obvious fix.
Wistia's State of Video reporting is consistent on the length relationship: shorter videos hold a higher percentage of their audience. That is a strong argument for cutting rather than for producing better, and the curve tells you exactly where the cut belongs. Read it alongside engagement rate and view rate, since a video that holds attention and converts nobody is a different problem from one nobody finishes, and the two get confused constantly.
Using the curve without over-reading it
- Mind the sample. Under a few dozen views the curve is noise. Aggregate similar videos before drawing conclusions.
- Compare like with like. Length, placement, autoplay setting and audience all change the shape. A personalized video sent to one named customer and a public marketing clip are not the same measurement.
- Fix the biggest cliff first. The opening seconds hold more lost viewers than everything after them combined, so that is where the first edit pays.
Try Venta Capture on your own process
Build one flow for your highest-volume case type. Free, no credit card.