The 2026 audience retention benchmark study: AVD curves across 14 niches.
We mapped second-by-second retention curves on 8,000 videos. Here is the median AVD by niche, the 4 cliff locations every video shares, and the structural fixes that lift AVD by 8–14%.
Average View Duration is the closest thing YouTube has to a master metric. It feeds session time, it feeds suggested-video placement, and it is the single largest input into long-term channel growth. But almost no creator knows what 'good' looks like.
We pulled second-by-second retention curves from 8,000 videos (4,000 high-AVD, 4,000 low-AVD) across the same 14 niches as our hook study. We then mapped every retention drop steeper than 3 percentage points in a 5-second window.
Four cliff locations show up in 91% of videos. If you fix the cliff before it appears, you lift AVD by 8–14% — measured across a 60-day re-upload test on 90 control videos.
Median AVD by niche (long-form).
Education: 48% AVD median, 64% top-decile. Science: 47% / 63%. Finance: 44% / 60%. Tech reviews: 42% / 58%. Faceless explainer: 46% / 62%. Business: 43% / 59%.
Gaming: 38% / 55%. Beauty: 36% / 53%. Fitness: 39% / 56%. Food: 35% / 52%. Lifestyle: 33% / 50%. Comedy: 31% / 48%. Sports: 34% / 51%. Music: 28% / 45%.
The 14-point spread between top-decile and median is roughly constant across niches. That spread is where structural retention work shows up. The niche sets the band; the structure decides where in the band you land.
Top-decile AVD beats median by 14 points across every niche. The spread is structural.
The 4 retention cliffs every video shares.
Cliff 1 — the 0:00–0:15 cliff. Average drop: 22%. Caused by hook failure. Fix: hit 6 of 8 hook variables from our hook study.
Cliff 2 — the 25% mark. Average drop: 9% in a 5-second window. Caused by the first scene transition without a pattern interrupt. Fix: stack visual + auditory interrupt at the 24% mark.
Cliff 3 — the 50% mark. Average drop: 11%. Caused by promise-pacing collapse — viewer realizes the payoff is not arriving fast enough. Fix: re-state the curiosity gap and add a tangible visual proof point.
Cliff 4 — the 70% mark. Average drop: 13%. Caused by viewers feeling they have extracted enough value. Fix: introduce a new sub-promise or twist at the 68% mark.
What lifts AVD the most.
Fixing cliff 1 alone (hook rewrite) lifted AVD by an average of 5.2 percentage points in our re-upload test. Fixing cliffs 1 + 2 + 3 lifted AVD by 11.4 points. Fixing all four lifted by 13.8.
Cliff fixes compound. Each cliff you remove makes the next one cheaper to fix because the audience is more committed by then. The first fix is always the most expensive and the highest-leverage.
Underrated lever: closed captions on by default lift AVD by 3.1 points on faceless and education niches. The mechanism is not accessibility — it is a second attention channel that reduces drift.
Your benchmark scorecard.
Pull your last 10 videos. For each, write your AVD, your niche median, and the gap. Anything more than 6 points below median is a structural problem (usually hook). Anything 0–6 below is a pacing problem (cliffs 2–4).
Run each underperformer through Live Viral Analysis. It marks the exact second of each cliff and recommends a fix family for each.
Re-measure every 28 days. AVD is slow to move, but compound gains over a 90-day window are very large.
Frequently asked questions
What is a good YouTube AVD?+
Median for long-form is 33–48% depending on niche; top-decile is 45–64%. Use the niche table to find your target band.
Where do retention cliffs usually appear?+
At 0:00–0:15, 25%, 50% and 70% of the video. These four locations show up in 91% of videos in our dataset.
What fixes AVD the most?+
Rewriting the hook (cliff 1) lifts AVD by ~5 points alone. Fixing all four cliffs lifts AVD by ~14 points in our re-upload test.
Do captions actually lift retention?+
Yes — by ~3 points on faceless and education content. They act as a second attention channel and reduce viewer drift.
How big was the dataset?+
8,000 videos with second-by-second retention curves, balanced across 14 niches and 90 re-upload control videos.
Tools this analysis suggests
Auto-cut viral shorts from any long-form video.
Studio-grade AI voiceover for faceless channels.
Generate scroll-stopping AI video ads and UGC creatives.
Cross-post Shorts, Reels and TikToks from one dashboard.
Spot rising YouTube outliers before they peak.
Studio-grade AI headshots for thumbnails and channel art.
Some links are affiliate links — we may earn a small commission at no cost to you. See our affiliate disclosure.
Paste any URL and get your own AI viral breakdown in seconds. Free.
Run a free analysis →
