Click, Watch, Stay

Part One · The measurements

Before anyone watches a video on YouTube, the site records a chain of small events. A thumbnail that stays on screen for more than a second with at least half of it visible is counted by YouTube as an impression. Click-through rate is how often viewers click on a video after seeing its thumbnail and title, and YouTube says half of all channels and videos fall between 2 percent and 10 percent. After the click comes watch time. On June 12, 2026, YouTube's blog said that Jimmy Donaldson, who publishes as MrBeast, was the first individual creator to reach 500 million subscribers. This book asks how the craft behind his channel fits those measurements, and where the public record runs out.

The channel did not begin with large numbers. Donaldson uploaded his first video in February 2012, at 13. His Minecraft and other game clips reportedly drew about a thousand views each. By mid-2016, the channel held around 30,000 subscribers. In January 2017, he posted a video of himself counting to 100,000. The counting took about 40 hours, and he sped up some stretches to keep the upload under 24 hours, yet within days it had tens of thousands of views. He has said that his first 500 videos did not deserve a million views. That treats the early years as a long period of practice rather than a sudden breakthrough.

Part Two · Editing for the graph

The measurement every creator learns to watch is the retention graph. YouTube's help page says that dips mean viewers are abandoning or skipping at that point. Spikes mean people are watching, rewatching or sharing a part, and a gradual decline is common, since videos generally taper off. Average view duration is the average minutes watched among those who stayed to watch. Donaldson's habits read like answers to that graph. He has said the first ten seconds are what people undervalue the most. And he has said that a longer video needs a side story in the middle. These are his own habits, not YouTube figures.

He has also set a threshold of his own. In a 2021 interview, he said he told a friend who makes videos that a video with under 70 percent retention meant something was done wrong. That figure is his rule of thumb, not a YouTube standard, and one cutoff cannot mean the same thing for a two-minute clip and a thirty-minute one. One description of his usual format is a premise in the first 30 seconds, a promised finale and a runtime of 10 to 30 minutes. In March 2024, the journalist Taylor Lorenz described the style known as retention editing as loud sound effects, fast cuts, flashing lights and zero pauses. That month, Donaldson said he was slowing his videos down. His simpler rule is that if he is even slightly bored, he cuts it.

None of that editing works without raw material. Donaldson has said his videos are unscripted and the filming is run and gun, so the crew must over-film. In a May 2024 interview, he said that for some videos, the footage was so large that it would have taken a single human five lifetimes to watch. His first hire, he said in 2021, was an editor, because he was so particular about how he wanted a video to look. In a pipeline like this, editing is the stage where hours of material are reduced to the one story that the retention graph will test.

Part Three · Packaging and testing

The other half of the pipeline is packaging. In a YouTube interview reported in March 2024, the thumbnail designer, Chucky Appleby, said the team makes about 50 thumbnail and title concepts per video. It varies shirt colors, outfits and backgrounds into hundreds of versions. Donaldson's face has been in the thumbnails since 2019, and Donaldson has called it brand recognition. Appleby said viewers who trusted one video may click the next one when they see the same face. His rules are easy to state. A thumbnail should be easy to understand, because pictures crowded with detail often fail at small sizes. And it must keep its promise. In his words, trust is what makes people come back and click on your videos.

YouTube's own testing tool turns this into a process. A creator can compare up to three titles and thumbnails on one video, and different viewers see different versions. The winner is the version with the most watch time. Results are labeled Winner, Performed Same or Inconclusive. A small share of viewers is held out as a control group that sees only the default version, and their data is left out of the results. A test should finish within two weeks. In June 2024, YouTube's Rene Ritchie said the tool returns watch time rather than separate metrics, because too much focus on click-through rate can lead to clickbait that hurts retention.

That design choice changes how to read a thumbnail. A click-through rate rewards the promise of a video, while watch time also tests whether the promise was kept. Donaldson has made a related point about momentum: the longer people watch something, he said, the more likely they are to keep watching. A picture that over-promises may win the click and then lose the viewer seconds later, which is the pattern a watch-time test is built to catch. It also explains why Appleby ties thumbnails to trust, since a viewer who feels misled once may not give the next video a chance.

The best-known public result of this testing concerns Donaldson's mouth. Appleby described about 30 videos in which the closed-mouth version had higher watch time every time. He called the overall difference small, but the consistency convinced the team, and closed mouths became the team's standard from then on. A March 2026 article in The Ringer quoted Donaldson: every A/B test, less cringe expression won. By our arithmetic, if each result were a fair coin flip, thirty in a row would happen about once in a billion tries, because two to the thirtieth power is 1,073,741,824. That suggests the pattern was not luck, though the arithmetic assumes independent tests, no unreported tests with other outcomes, and a complete account. Treat it as a hint, not a measurement. But the reports gave no watch-time numbers, so the size of the effect is unknown.

Part Four · Scale and what is missing

Scale is the part of the method that cannot be copied from outside. One report said that by 2023, Donaldson employed over 250 people, from writers to editors to producers. In January 2023, he said he still approved every idea before it was filmed. In May 2024, he called himself technically the product and said it all falls apart if he does not show up. He also said he does not recommend that most people make videos like his, because they take a lot of team and a lot of money. What does carry over is the loop: publish, read the graph, change one thing, publish again.

A large audience can also be used as a tool. On October 25, 2019, Donaldson and Mark Rober, a former NASA engineer, launched #TeamTrees. The goal was to plant 20 million trees by 2020, one tree for each dollar given. The Arbor Day Foundation ran the planting with forestry partners. The campaign's site says that, using final project reports from the foundation's planting partners, an independent accounting firm validated the planting of over 24 million trees, and the site now lists more than 25 million. Here money is counted, assigned and audited by a third party. That is a different kind of measurement from a retention graph, but it shows the same habit of defining a number before chasing it.

The milestones are straightforward to date. Guinness World Records verified his channel at 112,193,139 subscribers on November 17, 2022, a record for an individual male on YouTube at that time, and the 500 million mark came on June 12, 2026. A record like this lasts only until another creator passes it. What is missing is the data behind the craft. The reports reviewed here give the methods and a few summaries, but not the retention curves or the size of each effect. In May 2025, Donaldson told Guinness his advice was to keep failing until you succeed. He added that he enjoys doing things people have never seen or done. It is a loop that leaves little public data beyond the videos.