How Diary of a CEO Mastered Podcast Clipping
- Diary of a CEO went from 100,000 YouTube subscribers to 17.5 million in roughly five years, with about 50 million monthly listeners, and finished 2025 as the world’s No. 2 podcast on Spotify.
- None of it is luck: Bartlett flags clip moments mid-interview, the best lines become thumbnails tested as Facebook ads (click-through up from ~2% to 20%), and the team runs 150+ experiments a quarter.
- The machine even killed its own ideas: doubling the daily clip output reduced reach, so they stopped. Test, own the channels, measure. Everything else is staffing.
Steven Bartlett decides what gets clipped before the interview is over. While a guest talks, he rests a thumb on a trackpad and flags the moments worth cutting, live. By the time an episode of Diary of a CEO is published, its best lines have already been turned into thumbnails and tested on real audiences as Facebook ads. That is not a podcast being clipped. That is podcast clipping run as an engineering discipline.
The moment is flagged before the episode ends
Bartlett described the method himself at an advertising event in Sydney: he marks strong moments on a trackpad during recording, his team uses AI to turn the best-performing lines into ad titles and thumbnails, and those get tested as Facebook ads before launch. His verdict: “This has changed the game”, with click-through rates lifting from about 2% to 20%.
The testing goes deeper than thumbnails. Episodes are screened with thousands of viewers before publication, with eye-movement tracking, and cut based on retention data. The company employs a “Head of Failure”, whose mandate Bartlett summarises as: “The goal is to get it wrong as many times as possible, because that’s how you get it right.”
The DOAC clip strategy: an operating system, not a channel
DOAC’s distribution is structured like a product org. There is the main channel, a dedicated clips channel publishing one 10-to-30-minute clip every day, a separate shorts channel, and a behind-the-scenes vlog. Grace Miller, the team’s head of failure and experimentation, has described running over 150 tests in a single quarter, down to hand-in-shot versus no-hand-in-shot thumbnails, with several A/B/C tests in the first 24 hours of a release.
The most instructive experiment is one they killed. The team tested doubling the clips channel to two clips a day, and reach went down, so they went back to one. Compare that with the industry default, posting more and hoping, and you see the difference between running distribution and doing content. Miller’s one-line summary of how the channel hit 10 million subscribers: “You reach 10 million subscribers by caring about every tiny detail.”
The machine extends upstream of the edit, too. FlightStory CEO Georgie Holt has described a house-built tool called Guest Radar, which scans hundreds of thousands of channels to spot people starting to over-perform before they are famous. Even the guest list is data.
The business the machine built
The output numbers are the argument. Around 100,000 YouTube subscribers turned into 10 million by April 2025, the second-largest podcast on YouTube behind Joe Rogan, and 17.5 million by mid-2026, with roughly 50 million monthly listeners and a finish as the world’s No. 2 podcast on Spotify for 2025. Parent company FlightStory now runs about 150 people on a reported $47 million of annual revenue, and Bartlett has said he turned down a nine-figure offer for the show. FlightStory’s CEO frames the thesis in one line: “the next billion-dollar media company will be built out of human IP”.
For balance: the machine has drawn scrutiny as well as subscribers. A December 2024 BBC World Service investigation found repeated, potentially harmful health claims across health-focused episodes, aired with little challenge, and UK regulators banned two ads featuring Bartlett over undisclosed financial interests. Distribution power cuts both ways; what it amplifies is an editorial choice.
What everyone else can copy, and what they cannot
The copyable part is the philosophy. Clips are not promo for the episode; clips are the discovery engine, tested like ads and published on channels the show owns, where every subscriber compounds. Even the money agrees: Joe Rogan’s current Spotify deal, reported at up to $250 million, dropped exclusivity, which put full episodes and clips back on YouTube and everywhere else.
The part most podcasts cannot copy is a 150-person org with a Head of Failure. That gap, between knowing the playbook and staffing it, is exactly what a managed clipping operation closes: a private team finding the moments, cutting daily, posting to accounts you own in the markets your audience lives in, and attributing which clips actually turn viewers into listeners and listeners into customers. Views are not created equal, and a podcast’s clip machine should know which clips make listeners and which just make noise.
The DOAC playbook, without hiring 150 people
We run private clipping operations for podcasts and founders: your episodes, cut and posted daily on accounts you own, tested and attributed the way DOAC does it. The system in this article, staffed for you.
Frequently asked questions
Bartlett flags strong moments on a trackpad while recording, the team turns the best lines into titles and thumbnails, and they are tested as Facebook ads before launch, a method he says lifted click-through from about 2% to 20%. Episodes are also pre-tested with thousands of viewers, including eye-movement tracking, and the team has described running over 150 experiments in a single quarter.
DOAC runs a dedicated clips channel publishing one 10-to-30-minute clip per day, plus a separate shorts channel and platform-specific cutdowns. Notably, the team tested doubling to two clips a day and reach went down, so they killed the experiment. No official clips-per-episode count is published.
The principles copy: treat clips as the discovery engine, test before you publish, own the channels the clips live on, and measure what converts listeners rather than what merely gets views. What does not copy easily is the 150-person operation behind it, which is the gap a managed clipping team fills.
Flag the strongest moments while recording, cut them into clips built around one hook, test titles and thumbnails before publishing, post daily to dedicated channels you own, and keep only what the retention data supports. DOAC layers a 150-person operation on that loop; a managed clipping team gives smaller shows the same system without the headcount.
Sources
- B&T, Bartlett on live clip-flagging and Facebook-ad thumbnail testing (April 2024)
- Colin & Samir, recap of Bartlett’s Press Publish talk (September 2025)
- The Nathan Barry Show, with DOAC’s Grace Miller (March 2026)
- Dain Walker, interview with Grace Miller on DOAC’s experimentation cadence
- stevenbartlett.com, on the 10 million subscriber milestone (April 2025)
- RTE, on DOAC’s 2026 scale and the ASA rulings (Adam Maguire, June 2026)
- The Next Web, on FlightStory’s size, revenue and US expansion (July 2026)
- Business Leader, “Inside FlightStory” (January 2026)
- Lindsey Gamble, interview with FlightStory CEO Georgie Holt
- Pharmaphorum, on the BBC World Service investigation (December 2024)
- AP via NY1, on Joe Rogan’s reported $250M Spotify deal (February 2024)
- Wikipedia, The Diary of a CEO, scale and chart positions