How Fitbod grows

A near-constant in-app banner lets subscribers 'share six free workouts': the normal trial is three free workouts, while a friend who arrives via the shared link gets six. Built primarily to measure the word-of-mouth loop, with every step instrumented: % of subscribers who view the referral feature, % who send, % of recipients who tap, redeem and convert. Next step is to find the highest-leverage step (e.g. more views of the referral page, or an incentive for the referred user to convert).

Fitbod lets new users open the app and work out with no trial signup: three free workouts with every feature unlocked. After the first workout it says 'You've got two more workouts for free', then asks for a subscription.

The two co-founders spent the first several years alone building product and listening to customers (users emailed unprompted with feedback and requests), relying on engaged users to spread the word. Monetized in 2016 and only raised a seed round about 1.5 years later. Paid advertising was turned on only once conversion and retention (LTV) were strong, and treated as an accelerant to the word-of-mouth loop rather than the foundation.

Designed the product hook on Nir Eyal's Hooked model (trigger, action, reward, investment): trigger is uncertainty about what to do in the gym, action is opening the app for your next workout, and after logging a workout the app celebrates it with achievements and progress signals. The reward is made visible and shareable, e.g. a share screen showing total pounds lifted in the workout, so users show the value they get and prospective users see it.

In the early days the engineering co-founder (Allen) took an analytical approach to testing keywords and tracking rankings in Apple's search results.

During an early beta the app pivoted from an advanced workout tracker to recommending the next workout, using AI/ML to produce customized strength-training prescriptions ('only one workout: your next one').

Seeing the downside of grow-at-all-costs in an underdog category, picked a seed lead (Jason Calacanis) known to run against it, and filled later rounds mostly with startup operators aligned with the model, such as Calm's co-founders (who beat Headspace by staying lean) and a Notion operator who champions revenue per headcount.

Focus (the 'offense' side of Reforge's offense vs defense framework) is on converting ICP trial users who get far but don't subscribe. Quantitatively, compare the non-converter cohort's profile and trial actions with successful subscribers. Qualitatively: build a cohort in Mixpanel (e.g. people who logged workout two or three of their trial), export CSV, upload to userinterviews.com (free when you bring your own users; it handles scheduling and payment), with plain, authentic, high-volume invites. Run targeted 30-minute interviews: what problem were they trying to solve, and what actual or perceived gap is there to the product (perceived gaps are framing problems). Use the Five Whys to find the most constraining problem, then show a feature pitch or mock-up and ask 'If this feature existed today, would you buy Fitbod for $79.99?' A no means the real problem hasn't been found; understanding a feature is not the same as it solving their problem.

Retention is tracked at three levels: (1) subscription retention (did the subscriber auto-renew; takes a year to read on annual plans and includes inactive renewers), (2) cancellation retention (did they hit cancel during the period; readable monthly), and (3) workout retention, the metric they actually operate on: a subscriber must log 3+ workouts in a month to count as retained into the next month. The threshold is adjustable in the dashboard (1+, 2+, 3+, 4+) and informed by the ICP, who log about 2+ workouts a week.

Once interviews give a medium-confidence pattern (e.g. non-converters couldn't find a workout format or exercise), form a hypothesis and test it as cheaply as possible: an automated survey after the first three workouts ('How did your three workouts go? Was there an exercise or format you couldn't find?'), or better, an intervention on the one or two surfaces where users would look, asking 'Could you not find this, yes or no?' to size the problem before building.

Rather than trying to revive churned users (lower leverage), a model predicts user engagement, with logged workouts as the best predictor, to pinpoint users at risk of going dormant; engagement tactics and interventions are aimed there. Research focuses on why people go dormant rather than on features super-users request.

Investing 'quite a bit' in the company blog as a second growth loop alongside word of mouth: content drives new users, who produce revenue, which funds more content. User-generated content is also being considered.