The 70% Retention Myth Is Killing Your Channel

Mario Joos, CEO of BrightRock, started a research firm specifically so he could ask the questions that corporate content firms wouldn’t fund — since there was no guarantee of ROI. Joos spent several years working inside the MrBeast organization, left to share his knowledge with others who wanted to create similar types of content, and has subsequently worked with a number of creators whose collective reach exceeds 100 million. Everything Mario says about retention isn’t some kind of opinion. His views regarding retention are grounded in behavioral research dating back to the 1950s — plus a whole bunch of science about how our brains work, and lots of data related to videos that have had well over 100 million views.

In short, fast cutting isn’t going to help you. The 70 percent retention rate is pure fantasy. Plus, the difference between creating a channel that has 2 million views versus a channel that has 100 million views has absolutely nothing to do with how quickly you cut.

Retention Is Not a Percentage

Generally speaking, most creators are looking at their retention dashboard thinking it is some kind of gradebook. Good video = high percentage. Bad video = low percentage. Speed up the cuts. Fix the pacing. Problem solved. That type of reasoning is fundamentally flawed — and leads to a whole lot of wasted time and money.

According to Joos, the YouTube algorithm doesn’t just take your video and send it out to a completely random group of people and measure their response. Instead, it tries to determine which people are most likely to want to consume your video before even showing it to them. The retention percentage that you see isn’t reflective of some large test group. Rather, it reflects how the group of people that YouTube previously predicted would like your video did behave.

“I’ve worked on videos with over 100 million views that have a retention rate of 40%. When everyone advises you to get to 70%, that’s probably the biggest misconception. There’s no magic percentage. Just look at the behavior.“

What matters more than any of those metrics is something referred to by a Google engineering paper published in 2018-2019 as expected watch time per impression. Each time the algorithm determines whether or not to display your video to a particular viewer, it calculates: how much total watch time am I likely to receive from this single impression? For example, if you were to show your video ten separate times to the same viewer, and each time they viewed it for two hours, you would have accumulated much more value than if you showed it once and the viewer bounced off after thirty seconds. That’s basically the entire game — not trying to chase some arbitrary percentage on a dashboard.

The Three Components of Retention (Most Creators Focus on One)

Joos believes that every video contains three components: grabbing attention initially; keeping that attention until the end; and giving satisfaction — both during and after the viewing process. Properly defined, retention strategy involves everything that happens after the hook grabs someone’s attention.

The restaurant analogy he used is worth spending some quality time on. Let’s say you decide to visit a local steakhouse and order a nice piece of meat. The bread they give you right away isn’t the reason you decided to go there. However, if the bread is bad, it sets an awful expectation for everything else. Similarly, while your introductory sequence may matter — it’s not because it’s the main course, but because it sets expectations for what’s going to follow. Most creators spend their lives optimizing the bread sequence (i.e., the intro) and then serve a subpar steak.

Only focusing on optimizing the first 30 seconds while neglecting pacing, clear structural flow, character development and varied storytelling throughout the remainder of the video is essentially why creators don’t grow.

The Most Common Video Creation Mistakes (Regardless of Channel Size or Experience Level)

When asked by the audience about universal video creation and editing errors regardless of channel size, niche or experience level, Joos went into detail about how there were only two common mistakes he saw across all levels.

Non-Progressive vs. Progressive Content

Every single cut, every single line, every single action in a video, is either moving towards the end goal of the video or it’s not. A joke is non-progressive — it does not further the story; however, this is not always an issue unless the three jokes are stacked together. As such, pacing can slow down due to a lack of progress throughout the video.

In addition, this changes the way you will think about the overall editing process. Instead of asking “is my cut fast enough?” you should ask yourself “how quickly am I providing new progressive material that allows the viewer to want to continue watching.” Momentum is the key element here.

Too Much Repetition Equals Boredom

There is nothing worse than seeing something repeatedly. Whether it be visually, auditory or structurally, repetition kills attention. So, according to Joos, take any word associated with variety and add “variety” to the end of it. Then evaluate based upon those parameters. This includes visual variety, auditory variety, character variety and so forth. There are many times when a creator does not realize they have repeated themselves because they are reviewing their own work as opposed to the work that has been seen by the viewer prior to viewing the current piece.

Your Viewer Needs to Know Where They Are in Your Video

According to Joos, one of his favorite things to do during video production is make sure the viewer always knows exactly where they are within the video. Not necessarily with YouTube chapters, but with some sort of visual cue (title screen, segmentation graphic, etc.) that helps answer both questions: “where are we in the story?” and “how much time remains until we finish the video?” The easier it is for the viewer to understand where they are and what they need to do next, the easier it will be for them to remain engaged. When viewers become lost or confused, they often stop paying attention altogether. These are silent retention killers that may not appear directly in analytics reports.

Why Building a Short-Form Audience Won’t Automatically Help Your Long-Form Channel

Many creators believe that creating a large short-form audience will help create a larger long-form channel. Unfortunately, this is largely a misconception. Joos has access to the data and did not sugar coat it.

Direct conversion — i.e., when viewers click on a link within a Short to go view a long-form video — ranges from 0.2% to 0.5%. Based on 10,000 Shorts views, this means approximately 50 people clicked a link to view a long-form video. Similarly poor is indirect conversion — through the algorithm instead of a direct click. Together, direct and indirect conversions amount to roughly one long-form view for every 100 Shorts views at scale.

While this is not zero and can still add up if you are getting huge Shorts numbers, it is far from the flywheel that most creators assume exists. More importantly, while both formats reach the same individual, short and long form represent different media types used to meet different mentalities (or mind states) of individuals. An individual who is scrolling Shorts is in scroll mode — they are probably on their phone waiting somewhere (e.g., a doctor’s office), or perhaps they just got off a plane and are killing 20 minutes before a meeting. An individual who is watching long-form videos is likely sitting comfortably (and/or settled in) prepared to invest several hours of their time in viewing content.

The Basics of Creating a Parasocial Relationship Haven’t Changed Much Since Radio Was King

Joos’s reference to BrightRock’s parasocial relationship study from the 1950s demonstrates that there hasn’t been much of a paradigm shift in how we create those connections in seventy years. The way to create a parasocial relationship requires four components: intimacy, simulated reciprocity, persona, and repeat exposure. These components remain the same whether creating a parasocial relationship for radio or television audiences or now for your YouTube community.

What Creates the Connection in a Parasocial Relationship?

A number of creators believe that physical attractiveness will create a strong parasocial relationship (the Instagram model assumption). However, according to BrightRock’s research, this is simply not true. Two components drive parasocial relationships:

1. Social attraction (how likable, relatable and connected a creator is perceived to be)
2. Task attractiveness (how effective a viewer perceives the creator to be in their field/industry)

And physical attraction ranks a very poor third. This helps explain why so many large-scale followings created based on aesthetics do little-to-nothing to get followers to attend live events. Creators optimize for the wrong variables.

Likability and perceived competence are what allow viewers to establish a genuine emotional connection (enough to go to a live event, purchase products, etc.).

In addition to the above, there is another aspect of connecting with an audience that Joos discusses. Joos refers to a specific technique which involves switching between “YouTube mode” and “narrator mode.”

• YouTube mode: high energy, performed excitement, talking to the audience.
• Narrator mode: lower register, more personal, talking with the viewer.

The voiceover is typically where a creator will switch into narrator mode. For example: “this video took forever to make, thank you for watching.” This signal of intimacy is subtle but clearly demonstrated in various research studies related to establishing parasocial relationships. Apparently even Meryl Streep utilizes this technique in her acting work.

Why Dopamine-Based Content Creation Is Misunderstood

Much of the advice for creating content using dopamine is based upon a false premise. The false premise is that dopamine is responsible for making us feel good, and therefore we should use it to create rapid rewards to elicit feelings of happiness in our audiences. According to Joos’s research, this is incorrect.

Dopamine does not fire when we receive a reward. Instead, it fires in anticipation of a potential reward — often long before we receive it. Furthermore, dopamine only fires once after receiving a reward if that reward exceeds our expectation. The purpose of dopamine is not to give us pleasure — it is to motivate us to continue to pursue behaviors that lead towards rewards. After all, we don’t stop pursuing rewards because we received them; instead, we pursue additional rewards due to the positive outcomes achieved. Dopamine only fires again if the next reward significantly exceeds the last reward.

The implications for developing a successful content creation strategy are huge. While delivering a series of consistent rewards may provide some level of motivation for viewing, maintaining anticipation for future rewards is the key to keeping viewers engaged. Therefore, while the hook is designed to capture attention and entice viewers into continuing to watch past the first few seconds, its true purpose is to serve as an indicator that something greater-than-expected will occur later in the video.

The Gap Between 2M Viewers and 100M Viewers: A Deeper Issue

Joos hypothesizes about a slippery staircase with four individuals. He believes that the goal should be to edit between each person slipping, jumping, climbing over each other quickly. Each of those aspects (visuals, pacing, challenge) check the boxes for why someone would watch such a video. However, there is clearly one element missing.

The missing aspect is the depth. Who are these individuals? How do they each approach the challenge differently? Which strategy is one using vs. which strategy is the next individual using? What is the sequence of actions that allows the application of a strategy in real time, instead of a series of isolated single moments? Without the latter, the viewer will have fast cuts and will have no reason to care about the final result. The editing here is performing some form of work that should be done by the storyline.

The same principles apply to educational content. “10 tips to get fit — tip number one: drink more water” will lose the viewers almost instantly, since the tip was given to the viewers prior to any sense of curiosity developed, and the information provided is not new. On the other hand, “you need to drink water with salt, and there are five specific types of salt” establishes an interest in the topic because that is truly new and unknown information to the viewer.

“If you are looking at fast cuts, then you are looking in the wrong place. Ask yourself this question: can another smart person replicate this piece of content, without doing research and without thinking about it? If your answer to that is yes, then you probably shouldn’t be focusing on that area either.”

Joos gives a simple rule: look for the unseen areas where no other competitor is looking; do not assume that your competitors are stupid. Channels that have been stuck at 2 million views typically do well with the obvious things: pacing, thumbnail, hook. The channels that have gotten up to 100 million views are going further in depth on items that don’t show up when you run a quick content audit: character differentiation, strategic sequence, true novelty in information or format, and emotional stakes that make an outcome worth watching.

On AI: The Real Risk Isn’t Deepfakes

Joos wants to avoid taking a hot take on AI for the sake of it. His real stance is much more nuanced and worth spending serious time thinking about. As a tool — for iteration, structuring, flagging weaknesses in a script, generating options — AI is unequivocally useful. The concern isn’t that AI could create photorealistic content. The concern is structural.

When AI-generated content floods a platform at near-zero cost, it competes for the same impressions, algorithmic attention, and ad dollars as human-created content that requires many times more time and resources to create than AI-compatible content does. Over time, those diverted resources put pressure on human creators to create content that fits patterns compatible with AI rather than content that requires creative investment. Joos calls this “filtered creativity” — a slow narrowing of the kinds of work done by humans to survive economically.

The second problem is systemic: when platforms are overwhelmed with low-effort AI content, the ability of algorithms to tell high-quality content from AI-generated content degrades. This is both a creative problem and a drain on the systems that allow discovery to happen at all.

He doesn’t want to ban AI — he wants to ask harder questions: what will happen to human creativity if the economic conditions supporting it are gradually eroded? That is a question without easy answers — and most people in the creator economy aren’t even asking yet.