How VidCC analyses YouTube retention drops
VidCC aligns your YouTube retention curve with the video and audio, then reviews what happens around changes in the curve. It describes the moment, offers a possible explanation and suggests an edit. High, medium and low priority help you decide what to review first.
What goes into an analysis
In the customer workspace, an analysis of your connected channel needs a public video, its duration and usable retention data from YouTube. The video title and available description provide context. Average view duration and average percentage viewed can add context when YouTube supplies them.
VidCC maps the curve's positions to the video timeline using the video's duration. The analysis receives the curve alongside the video, so it can connect a change in viewing with what is visible and audible around that time.
Retention shows the change
The measured curve helps locate drops, flatter stretches and spikes. It describes viewing activity across the video. It does not report a viewer's reason for leaving.
Video supplies the event
The analysis reviews what happens on screen, such as a demonstration, a repeated setup or a transition. A finding should describe an event that actually happens at its timestamp.
Audio supplies the spoken and audible context
Speech and sound help explain the moment. The analysis can flag an audible problem, such as distracting music or unclear speech, separately from retention findings. An audio finding does not require a drop in the curve.
Pacing describes how the video progresses
The analysis considers when useful information arrives, whether an explanation repeats, how sections connect and whether viewers have time to understand an example. Pacing recommendations come from interpreting the video and audio. VidCC does not calculate a separate measured pacing score.
How the curve guides the review
A change in the curve gives the analysis a place to investigate. The footage and audio help it propose an explanation and an edit. The same shape can have different meanings in different videos.
- Early drop
- Inspect whether the opening delivers the topic promised by the title, gives necessary context and reaches useful content. A fall in the opening alone does not prove the intro caused viewers to leave.
- Plateau
- Inspect what the video shows and explains while the curve stays steadier. It may reveal something worth preserving, but a flat line alone does not prove satisfaction.
- Slow decline
- Look at local changes alongside the surrounding trend. Loss spread across a whole video should not automatically become one severe finding. Leaving after a completed answer can also be natural.
- Spike
- Check for a replayed explanation, a moment people seek to or a section that is hard to follow. A spike alone does not establish that a moment is successful.
How VidCC sets drop priority
The AI returns a high, medium or low label with a retention finding, its timestamp and the suggested edit. The label helps you choose what to review first. It is not a separate numerical score calculated by the chart, a measured cause of viewer departure or a predicted improvement from making the edit.
Analysis versions can use different priority rules. The meanings below describe VidCC's contextual policy; earlier versions can use fixed curve thresholds. The contextual policy asks the analysis to consider the local retention change, how long it lasts, the nearby trend, the available data and the importance of the observed content problem.
High priority
A substantial content problem, supported by strong local retention evidence. Review the moment and proposed fix first.
Medium priority
A credible problem that matters to the video's experience, with supporting retention evidence. Check whether the proposed edit would improve that section.
Low priority
A smaller improvement opportunity, or one with uncertain or absent retention association. Use the observation to guide your own review.
Why drop size needs context
For an illustrative calculation, a curve falling from 50% to 40% changes by 10 percentage points. Compared with its earlier level, that is a 20% proportional decline. Neither number proves that 20% of unique remaining viewers left, because retention also reflects seeking and repeat viewing. This calculation is also different from YouTube's relative-retention comparison with other videos of similar length.
The size of a change is only part of the review. A sharp change over a short stretch needs different context from the same change spread across several minutes. Sparse samples also limit how precisely a change can be located.
A percentage change alone does not define a universal priority label. In the workspace, moments appear in time order so you can review the video as it unfolds.
What to inspect in the first 30 seconds
Our long-form retention framework focuses on the video's promise, useful progression and how easily the viewer can follow it. For the opening, that gives you specific review questions. These are things to inspect, not a ranking of the most common causes of early drops.
Does the promised content arrive?
Check when the opening first shows or explains what the title offers. Consider whether greetings, channel updates or setup delay the reason someone clicked.
Does each sentence move the video forward?
Look for repeated promises or explanations that add no useful information. A recap can still be necessary when it helps viewers understand what comes next.
Can the viewer follow the explanation?
Check whether an example appears when it is needed. Several unfamiliar rules before any demonstration may be harder to follow than an explanation tied to an action on screen.
Is the important speech audible?
Listen for distracting sound or unclear speech around the moment. Treat an audible problem as something to review even if the curve does not show a matching dip.
Worked example: an opening with too much setup
Imagine a tutorial whose title promises a camera setting that fixes flicker. Its first 25 seconds introduce the channel and list five technical rules. The first demonstration appears at 00:26, while the sampled retention curve falls through the opening. This is an invented example to explain the method.
Describe the observation
The opening lists rules before showing the flicker problem or its fix. The curve declines during that stretch. These are two observations about the same time range.
Offer a possible explanation
Viewers may be waiting for the promised demonstration or struggling to apply rules they have not seen in use. The curve cannot establish which explanation is correct.
Suggest a concrete edit
Try opening with the flickering shot beside the fixed one, name the setting, then introduce each rule when the demonstration needs it. This is an edit to test, with no promised retention gain.
How findings can be checked
VidCC's evaluation tools support human comparisons of analysis outputs and checks against reviewed timestamps. Human comparisons ask which analysis would help the creator more. Reviewers can check whether a described event is present and whether the suggested edit fits the video. Timestamp checks ask whether the analysis found reviewed moments and how close its timestamps were, within a declared tolerance.
Finding the right timestamp and explaining the moment well are different questions. A timestamp check does not prove that an explanation is correct, and reviewer preference does not establish that an edit will improve retention.
What the analysis can and cannot establish
YouTube's Analytics API returns retention at sampled intervals tied to the video's length. Its documentation describes 100 points per video. On a two-minute video the intervals are 1.2 seconds; on a two-hour video they are 72 seconds. A precise visible event can still have an uncertain association with a change between curve samples.
Seeking, repeat viewing and differences in who watched can affect the curve. A matching timestamp is evidence to investigate rather than proof of why an individual viewer left.
VidCC can be wrong, and no edit guarantees more views. Watch the flagged moment yourself. Test one change in your next comparable video and review its curve, while allowing for differences in audience and topic.
See it on your own videos
Connect your channel and choose a public video with usable retention data. VidCC shows the curve beside the playable video and marks moments to review. Run an analysis to get observations and suggested edits, then ask Viddy about a finding when you want to explore an alternative.