Built With Science

Members' demand and the need for personalization pushed the courses into an app. The channel started in 2018 and about 100,000 people joined its web courses, tracking workouts in phone spreadsheets and nutrition in a separate tool. Users "wouldn't stop bothering us about an app," and a more personal product needed to leave spreadsheets behind.

At the time the channel had about 1 million subscribers and was adding about 1 million a year.

Sub Club

A big audience gives a launch spike and a baseline, not a growth engine. The app launched in January 2025 to a spike from about 3 million subscribers, and older videos now bring a steady number of trials each month.

That ceiling is why the team built paid channels that do not depend on the channel itself.

Sub Club

Fix trial-to-paid and retention before buying ads, since paid traffic converts lower. They ran no Meta ads until January 2026, spending the first year on product changes driven by launch feedback, and only started ads once trial-to-paid, retention and churn looked right. Meta then scaled from zero to $100,000 a month in profitable spend.

Organic YouTube traffic is warm and converts at 35 to 40% trial-to-paid, against about 25% on Meta. "As long as we're constantly improving our trial-to-paid and retention, that lifts the floor for everything else." Revenue and payback on the ad spend were not shared.

Sub Club▶ 17:22

A long personalized web quiz converts cold traffic better than the App Store. About 90% of traffic goes to a long quiz on the website that educates, overcomes objections, raises perceived value and builds the user's plan from the answers. In head-to-head tests quiz traffic converted much better than direct App Store traffic.

The same quiz rebuilt almost one-for-one in the app still converted worse than the web, while the Android version converted like the web. No numbers were given, and App Store visitors may have higher intent, so a retest is being considered. One hypothesis is that app users want a fast solution while web visitors are researching.

Sub Club▶ 25:15

A discounted gym buddy add-on in the quiz lifts order value and retention. The quiz asks whether users want to add a gym buddy, with 15% off for both, and around 15% of everyone who starts a trial adds one. Partners also keep users subscribed longer.

It came from members asking for a family plan to bring a husband or wife, and a product manager proposed testing it first as a quiz option. Users who decline only see a small toggle on the pricing page, keeping the paywall simple. The retention gain was stated without data.

Sub Club▶ 30:00

Free calculators that feed the quiz recover search traffic lost to AI. SEO, about 30 to 40% of traffic, began declining as people moved to AI search, so they hired a consultant, optimized existing content and built calorie and macro calculators that lead into the quiz. The tools rank well in AI search and convert slightly higher than blog posts.

Sub Club▶ 36:45

Selling a coach, not a tracker, has held the $189 price since launch. The plan costs about $189 a year or $30 a month, unchanged since the 2025 launch, framed as a personal trainer in your pocket so buyers compare it to a human trainer. Meta ads never call it a tracker.

"You're comparing costs of a personal trainer versus a trainer in your pocket." The team prefers fixing the funnel over touching price.

Sub Club▶ 28:00

An annual-only paywall with a trial timeline moved buyers to annual. The variant showed only annual, with a Blinkist-style timeline: day 1 your personalized plan, day 7 the plan adapts, day 12 a reminder before the trial ends. Annual share rose from about 60% to 75 to 80% with no price change.

Monthly moved behind a view-all-plans option. "We made the risk of actually starting the trial a lot more clear."

Sub Club▶ 54:00

A losing no-trial toggle won once both options sounded good. A toggle offering 20% off for paying upfront framed it as clearly better and the trial as a loss, and it tested negative even with a 30-day money-back guarantee. A new arm framing both options as pleasant has shown a very strong lift over a few weeks.

"We make both options seem pleasant. So, it's not one framed as good or bad."

Sub Club▶ 57:30

More explanation on the paywall hurt buyers who were already motivated. Adding context to a pricing page to clarify the value proposition converted worse than having no added context at all.

The team's read: users already excited to buy get overwhelmed by more information at that moment.

Sub Club▶ 55:20

Ads use B-roll, testimonials and AI statics to solve the bus problem. Jeremy's time is the scarce resource, so ads use almost none of his content beyond a few old shorts, and his time stays on long-form YouTube, the highest-ROI work. The goal is a channel "that can scale on its own."

The team admits his face would likely make ads perform well. Outside clippers post long-form cuts to Reels and TikTok, and viral clips show which hooks to use in future videos.

Sub Club▶ 14:45

Target specific avatars and pain points, because cheap click-bait trials don't pay. A month of Meta data set a baseline trial-to-paid and the allowable acquisition cost; then they tested avatars built from customer surveys and warm traffic and iterated on winners. Click-bait ads brought cheap trial starts but trial-to-paid dropped sharply.

"We want people to start the trial at the cheapest cost, but also we don't want just anyone to start the trial." An engineer built a dashboard confirming which trials truly came from Meta, and the highest-intent users are fed back to Meta.

Sub Club▶ 22:40

After Andromeda, creative does the targeting, not Ads Manager settings. Much of Meta's targeting now happens through the creative, so they spend less time on bids and settings and more on making content people enjoy, as on YouTube.

Sub Club▶ 24:40

Run Meta in-house so ads promise only what the product does. They chose in-house management over an agency to get someone who knows and likes the product and won't over-promise, for a seamless handoff from ad to app.

"Promote the app in a way that isn't faking anything."

Sub Club▶ 21:30

Story videos about a user's transformation drive more trials than broad education. They track nearly every click from videos and description links, so each video's trial yield is visible. Broad diet or program videos reach more people but activate fewer trials than stories of a user transforming on the app.

New content is planned to tie a story into the product pitch. The framing is "how does the app complement this piece?" rather than which piece pitches the app.

Sub Club▶ 6:45

Soft mid-video calls to action and expert guests protect trust. Most videos mention the app in the middle or near the end with a smooth transition, keeping full pitches for formats like Instagram shorts. Jeremy collaborates with lesser-known researchers, coaches and dietitians to add credibility.

The team plans to double down on trust because AI makes it possible to build almost anything. Saturation depends on the hook: a viewer bored by a diet video may be hooked by a workout video, and the algorithm then surfaces other videos as a mini funnel.

Sub Club▶ 6:45

Agree on the problem before brainstorming solutions, and write it down. They find the biggest drop-off in Amplitude, list every reason the page might fail, such as lack of trust or an unclear value proposition, agree on the biggest one, then list fixes. About 70% of experiments fail first, and the notes let them diagnose and iterate.

"We're obsessed with the problem and not the solution." The no-trial toggle is an example of a failure turned into a win.

Sub Club▶ 37:43

With limited traffic, test one big batched swing per problem. Instead of many small tests, they bundle changes that address one problem and strip the batch down if it fails. They skip colors and small text tweaks for high-leverage funnel drop-offs.

"We won't have enough to do a bunch of mini tests. So we'd rather focus on a bigger swing."

Sub Club▶ 42:45

Jeremy AI, built in a weekend, became a customer favorite. One engineer built a chat trained on Jeremy's YouTube videos and blog history that also reads each user's workout, nutrition and progress data for personalized answers.

Sub Club▶ 46:30

Launch features silently to read real engagement before announcing. Emailing all members about new features inflated engagement, so they now ship without saying anything, watch the metrics and announce only if there is signal.

Beta pools of about 100 long-time members and about 1,000 testers hear first and give feedback before the quiet launch.

Sub Club▶ 44:48