The YouTube Retention Glossary

Every term you will meet in a retention graph, a Studio report, or a Retti analysis — defined precisely, with a practical note on what to do about each one.

Updated June 2026 · Definitions reflect how these terms are used in YouTube Studio and in Retti's retention analysis.

Audience retention

Audience retention is the percentage of a video that viewers watch before leaving, measured by YouTube for every public upload. It is reported as a single average for the whole video and as a second-by-second curve. Retention measures how well a video holds the attention it has already won; it says nothing about how many people clicked.

Find it in YouTube Studio under a video’s Engagement tab. Work from the curve rather than the average: the average hides exactly where viewers left, which is the information you can act on. Full guide: /guides/what-is-audience-retention.

Retention curve

The retention curve is the line graph YouTube draws for every video, showing the percentage of viewers still watching at each moment of the runtime. It starts near 100% and falls as people leave. Its shape — cliffs, steps, plateaus, and spikes — is a second-by-second record of where the video held attention and where it lost it.

Read shape before height. A step down marks a specific mistake at a specific timestamp; a smooth slope is ordinary decay. For the overwhelming majority of videos, the steepest per-second fall sits right at the start, inside the first tenth of the runtime. Full guide: /guides/retention-curve-shapes.

Absolute retention

Absolute audience retention shows the raw percentage of viewers still watching at every moment of one video, with no comparison to anything else. This is the default graph in YouTube Studio: it begins near 100% and declines across the runtime. It answers a single question — of the people who started this video, how many are still here?

Use the absolute view to find drop timestamps in your own video, then judge the heights against real-world medians: most videos hold a clear majority of viewers at 30 seconds and well under half by the midpoint. Benchmarks: /guides/what-is-a-good-retention-rate.

Relative retention

Relative audience retention compares your video’s moment-by-moment hold on viewers against other YouTube videos of similar length. Rather than a percentage, it reports whether each part of your video retains viewers better or worse than is typical. A below-average stretch means viewers leave there faster than they leave comparable videos at the same point.

Relative retention is good at exposing weak sections you might excuse on your own graph, because the comparison strips out normal decay. It is a ranking, not a count — it cannot tell you how many viewers remain, so read it alongside the absolute curve rather than instead of it.

Average view duration (AVD)

Average view duration is the mean time viewers spend watching a video, shown in minutes and seconds. YouTube calculates it as total watch time divided by total plays, counting replays and early exits alike. AVD rises with stronger retention and with longer runtime, which makes it a measure of attention held in absolute time rather than in proportion.

AVD sits on every video’s analytics overview in Studio. Because a longer video can post a higher AVD while keeping a smaller share of its audience, never compare AVD across videos of different lengths without also checking average percentage viewed. Full guide: /guides/what-is-youtube-avd.

Average percentage viewed (APV)

Average percentage viewed is the share of a video’s runtime the average viewer watches, expressed as a percentage. It is the proportional twin of average view duration: AVD measures minutes, APV measures the fraction of the video those minutes represent. A 10-minute video with a 4-minute AVD has an APV of 40%.

APV makes videos of different lengths comparable, which AVD alone cannot do. Expect it to drift downward as runtime grows even when videos are performing well, so judge it against similar-length uploads. For choosing between the two metrics: /guides/avd-vs-apv.

Watch time

Watch time is the total number of hours viewers have spent watching a video or a channel, summed across every view. It combines audience size with holding power: more viewers watching for longer produces more watch time. YouTube reports it per video, per channel, and over any date range, and treats it as a core measure of the attention content earns.

Watch time is the volume metric; retention is the quality metric underneath it. When watch time falls, work out whether views dropped, retention dropped, or both, because the fixes are different. You will find it on the Studio dashboard and in the Content tab for each upload.

Hook

The hook is the opening stretch of a video — roughly the first 30 seconds — whose job is to confirm the viewer’s click and give them a reason to stay. A working hook delivers on the title immediately and sets up what comes next. It is judged by 30-second retention, not by how clever it sounds.

This is where videos lose the most: most videos keep only a fraction of their viewers by 30 seconds, and a large share have already shed a big chunk of their audience by then. Open with the title’s promise, not a channel introduction. Full guide: /guides/first-sentence-hook-data.

Opening cliff

The opening cliff is the steep fall at the start of nearly every retention curve, produced by viewers who click, sample a few seconds, and leave. Part of it is unavoidable mismatch traffic. Its depth is set by how quickly the opening proves the video matches the title that earned the click.

You cannot remove the cliff, only shrink it. For the overwhelming majority of videos, the steepest per-second drop falls right at the start, inside the first tenth of the runtime, and most videos still hold a clear majority of viewers at 30 seconds. Compare your own figure before judging the rest of the video. Benchmarks: /guides/what-is-a-good-retention-rate.

Drop-off point

A drop-off point is a timestamp where the retention curve falls clearly faster than the surrounding slope — a step or cliff rather than gradual decay. It marks a moment where many viewers made the same decision to leave, and the trigger is almost always something that happened in the seconds just before it.

Find your three biggest drop-off points after the opening and watch the 20 seconds before each one; the cause lives there, not at the drop itself. Usual suspects include tangents, repeated points, and payoffs that still have not arrived. Full guide: /guides/why-viewers-stop-watching.

Rewatch spike

A rewatch spike is a rise in the retention curve: a moment registering more views than the moments before it. It appears when viewers rewind to replay a section or skip directly to it. Spikes mark the passages an audience found most valuable, most surprising, or — sometimes — hardest to follow on the first pass.

Genuine spikes are rare: only a small fraction of videos show a clear upward rise after the opening. When you earn one, study the moment and make more like it — but check it is not viewers rewinding a confusing explanation. Curve patterns: /guides/retention-curve-shapes.

Open loop

An open loop is a question a video raises and deliberately leaves unanswered for a stretch, giving viewers a concrete reason to keep watching until it closes. An unresolved problem, a promised reveal, or a result teased before its explanation all function as open loops. The technique fails when loops open and never pay off.

Run one main loop — the title’s promise — and open the next question shortly before closing the current one, so nothing is ever fully settled until the end. Viewers forgive a delayed answer; they resent a missing one. More structural techniques: /guides/improve-youtube-retention.

Payoff

The payoff is the moment a video delivers what its title and hook promised — the answer it set up, or the result it has been building towards. Every promise a video makes creates a debt, and the payoff clears it. Videos can carry one central payoff or stage several smaller ones along the way.

Late payoffs are a leading cause of mid-video decay: when viewers reach the midpoint without the first instalment of what they came for, the curve sags and keeps sagging. Stage partial payoffs early rather than saving everything for the end. Full guide: /guides/why-viewers-stop-watching.

Stakes

Stakes are whatever is currently at risk in a video: what could be lost or won, and why the outcome is uncertain. Stakes give viewers a reason to care about the next minute rather than just the conclusion. When nothing is at risk, watching becomes optional, and retention curves register that quickly.

Audit any flat section by asking what is uncertain right now. If the answer is nothing, add a constraint, a deadline, a cost of failure, or a prediction that might be publicly wrong — and restate the stakes after each payoff, because resolved stakes stop pulling.

Pattern interrupt

A pattern interrupt is a deliberate change in what the viewer sees or hears — a cut to b-roll, a graphic, a location change, a shift in music or pace — placed to break monotony before attention fades. It resets the viewer’s sense that something new is happening without changing the subject of the video.

Use interrupts against the slow sag that long, unbroken talking-head stretches leave on a retention curve. The change must be visible or audible, not merely a new sentence, and the type should vary — a repeated interrupt becomes a pattern itself. More techniques: /guides/improve-youtube-retention.

Pacing

Pacing is the rate at which a video delivers new information, visual change, and forward movement. Fast pacing means short gaps between new things; slow pacing means long ones. Good pacing is not maximum speed — it matches the delivery rate to what each section needs, accelerating through setup and slowing for moments that deserve weight.

Pacing problems show on a retention curve as long sections of steady, steeper-than-normal decline with no single cliff. The fix usually lives in the edit: tighten gaps, cut throat-clearing, and move the strongest material earlier rather than saving it.

End-screen drop

The end-screen drop is the fall in retention across a video’s closing seconds, as viewers leave once the content feels complete — typically when an outro, a recap, or the end screen itself signals that nothing new is coming. A modest drop here appears on almost every video and is not a content failure.

For most videos, only a small, normal amount of retention is shed over the final tenth of the runtime. A far larger late slide usually means the video outlived its idea. Cut the long goodbye and point the end screen at a specific next video. Curve shapes: /guides/retention-curve-shapes.

Session time

Session time is the total length of a viewer’s continuous YouTube visit, counted across every video they watch in one sitting. A video extends a session when viewers keep watching afterwards — your content or anyone else’s — and ends one when viewers leave the platform altogether.

Studio does not show session time directly, so treat it as a principle: end videos by handing viewers an obvious next step, such as a linked follow-up or a strong end-screen suggestion, instead of an outro that winds the energy down and invites them to leave.