The DOAC Machine: How Steven Bartlett Out-Tests Everyone

In 2024, Steven Bartlett turned down a streaming deal reportedly worth $100 million. His stated reason should be pinned above every creator’s desk: “We looked at what they did in terms of testing, experimentation, innovation, and I felt like I was looking at the past.”

A podcaster turned down nine figures because he believed his in-house testing operation was worth more. And the numbers suggest he was right: a billion total streams, roughly 50 million monthly listeners by the end of that year, and by late 2025 the fastest-growing podcast in the world — eventually overtaking Joe Rogan in his home market.

The part everyone skips

The Diary of a CEO did not start as a phenomenon. It started in September 2017 as Bartlett — a university dropout who’d built and left the Social Chain marketing agency — talking alone into a microphone about entrepreneurship. Five solo episodes before a single guest. Four years of compounding before the sponsorship revenue crossed seven figures. The overnight-success arc, as ever, was a half-decade long.

The machine

What separates DOAC isn’t interviewing talent — Bartlett says so himself:

“None of that is because I’m a good interviewer… honestly, it’s because we out-fail our competition.”

“Out-fail” is the whole philosophy: run more experiments than anyone else, and let the losses buy the wins. In practice — per a widely-cited analysis of their growth operation — that reportedly means on the order of a hundred thumbnail variants tested per episode, ads run before release day so an episode launches already knowing its best packaging, and a dedicated “Experimentation Manager” whose job is making sure every team is running a test at all times.

Read that again slowly: the video is optimized before it exists publicly. Most creators pick a thumbnail by gut at 11pm on upload night.

The indie version of the machine

You don’t have a data team. Fine — the principle scales down further than you’d think:

• Generate real packaging options. Ten title variants cost nothing; ten thumbnail concepts cost an hour with an AI generator. The failure mode isn’t picking wrong — it’s only ever making one.
• Test before you publish. A community-tab poll, a story vote, even texting two options to five honest friends beats zero data.
• Make experiments a habit, not an event. Bartlett’s crew institutionalized it. Your version: change exactly one packaging variable per upload and write down what happened. Twenty videos later you have something almost nobody has — evidence.

And his north-star metric is a relief, honestly: “The most important thing in podcasting isn’t actually the amount of followers someone has… it’s ultimately about the value that it gives, because people will share it.” The testing machine exists to package value, not to replace it.

CreatorSteals has no affiliation with Steven Bartlett or Diary of a CEO. Facts sourced from public reporting, linked throughout; thumbnail-testing figures are from third-party analysis of DOAC’s publicly described methods.