People-First YouTube Algorithm
Optimise discovery, attention, and satisfaction for real viewers
- Difficulty
- Advanced
- Time to result
- ~ongoing to results
- Steps
- 6
- Confidence
- 99%
The People-First YouTube Algorithm combines four main inputs: previous viewer behaviour, click-through response, watch time, and viewer satisfaction. YouTube first matches content to clusters of interests inferred from what people already watch, making a clear niche and relevant neighbouring channels useful. Topic, title, and thumbnail then compete for the click. Once viewers arrive, average duration, percentage viewed, retention, and continued viewing reveal whether the content holds attention. Finally, surveys, return behaviour, likes, comments, and bingeing help indicate satisfaction. No isolated metric is universal: click-through rate often falls as distribution broadens, while a long video can succeed with a lower completion percentage but high minutes watched. The durable rule is to replace “algorithm” with “people” and optimise the entire experience.
Origin
Sean Cannell organises YouTube's disclosed signals into four factors, then simplifies the system by telling creators to replace the word algorithm with people.
Core principles
- 01The algorithm models people rather than replacing them
- 02Viewer history creates niche recommendation clusters
- 03Topic, title, and thumbnail earn the click
- 04Watch time and satisfaction sustain distribution
- 05Returning and bingeing viewers signal durable value
How to run it
- 1
Map viewer behaviour
Identify the topics, channels, and adjacent interests the intended audience already consumes.
Pro tip Competition can enlarge the recommendation pool rather than merely crowd it.
Watch out Do not chase an unrelated trend that confuses the channel's audience associations.
- 2
Earn the click
Select a timely, relevant topic and package it with a compelling title and thumbnail.
Pro tip Treat topic as the foundation; packaging cannot rescue weak relevance indefinitely.
Watch out A high click-through rate on very few impressions is not proof of broad appeal.
- 3
Hold attention
Open for a new viewer, deliver efficiently, and use retention curves to identify dips and excess length.
Pro tip Add one extra round of revision to remove fluff.
Watch out Do not speak only to returning fans when the goal is cold discovery.
- 4
Measure minutes intelligently
Consider both average percentage viewed and absolute average duration in the context of video length.
Pro tip A strong long-form video can create exceptional total watch time despite modest completion.
Watch out Avoid applying one retention threshold to every format.
- 5
Deliver satisfaction
Pay off the title and leave viewers feeling that the session was worthwhile.
Pro tip Use comments, surveys, return viewers, and qualitative feedback together.
Watch out Retention tricks that delay a weak payoff can produce watch time but dissatisfaction.
- 6
Create continued viewing
Connect related videos with series, end cards, and a coherent library so viewers can keep solving the same problem.
Pro tip Track average videos viewed per viewer where available.
Watch out Do not force a next video that is unrelated to the viewer's current goal.
In the wild
A woman searching for marathon training begins receiving related shoe, nutrition, and coaching videos. A creator whose title earns her click, whose lesson helps her train, and whose library supports the next questions becomes a repeated recommendation for similar runners.
→ Discovery expands because the creator satisfies a real cluster of viewer needs rather than gaming one metric.
Common mistakes
Optimising one metric
A high click-through rate, completion rate, or view count has little meaning without impression scale, length, and satisfaction context.
Delaying a weak payoff
Keeping viewers waiting can raise watch time while making them regret the viewing session.
Copying a famous creator
Viewer-history matching rewards relevance within a niche, not imitation of the largest channel.
Is it for you?
Best for
It is best for creators using YouTube analytics to improve an educational or expertise-led channel over time.
Not ideal for
It is not ideal for anyone seeking a guaranteed threshold or shortcut independent of audience value.
From the transcript
“bottom line YouTube has two major metrics click through eight in watch time”
“the fourth one almost nobody knows about and it's viewer satisfaction”
“replace the word algorithm with people you're not optimizing for the algorithm you're optimizing for people”
From the episode
Sean Cannell: Start a Profitable YouTube Channel in 2024
Sean Cannell