"40% of all subscriptions take place after six months," says Mike Charniatsou, who works on growth and web onboarding at Flo, the women's health app. For most subscription businesses that number would be a problem to fix. At Flo it is the model. "We aim to be a lifelong partner to the user," Mike says, and that means accepting that a user's needs, and her willingness to pay, move with the stages of her life. "They may easily fall in and out of different needs and, you know, the level of willingness to pay, maybe different. And that's okay. So we accept those, you know, rules of the game."
Cycle tracking, the core use case, stays free. What Flo does instead is watch how each user's needs evolve and step in when it has something new to offer, whether that is trying to get pregnant, following a pregnancy, or understanding perimenopause. "When we see that there is something of value that we can bring to the user besides our core product offering, we do. And this is, you know, what lets us convert users from freemium" into paying, three, six, and 12 months later.
Mike is clear that this is a luxury, not universal advice. A younger app under pressure to pay back acquisition from day zero can't wait six months for its subscribers. But he also thinks the alternative has a ceiling. There are only so many ways to squeeze the lemon, he says, "and suddenly, you know, like there are not that many ways to keep squeezing that lemon." Flo would rather protect the relationship: "We don't feel the need to try and monetize the user like on day one and day seven, you know, because we know that, you know, one day something may change."
"Your paywall conversion is not your monetization strategy"
That line is Mike's, and David Barnard read it back to him on the show. The argument behind it: growth teams overestimate two things. "One of them is that short term conversion wins smoothly translates to business wins, which is not true." The other is that subscription conversion is a single event. "There are lots of things like starting a trial, activating a trial, you know, retaining the user and so forth. So it's a life cycle." An onboarding paywall is the beginning of that journey, and a test that lifts it can still lose money further down.
David's counterexample came from an earlier Sub Club guest, Anada Lakra of BoldVoice, whose team removed the X from its paywall, watched conversion jump, and then watched refunds spike until the win became a drain. Mike's answer is to keep analyzing experiments long after rollout, and to run one exercise at the end of every year. "Over the course of a year, you kind of ship tons of successful experiments, you know, with five, seven, 10% uplift, you know, and at the end of the year, you just, you know, do this exercise and check. Did your business grow by the same percentage that you've reported over the course of the year?" If it didn't, the difference is the share of wins that were never business wins.
The onboarding step that lost 10% twice and then won 23%
Mike's biggest win of the year started as a failure. Flo's web onboarding team added a step to work out which perimenopause phase a user was in. "It made so much sense to us at the very beginning. The hypothesis was strong," Mike says. "And the first test came back with almost 10% drop in conversion rate."
What happened next is the part he wants other teams to hear. Instead of looking for evidence, the team reached for an explanation. "Maybe something was, you know, something was wrong with the design. Let's just do it. One more run and nothing happened. You know, we got the same job and conversion rate." Only after the second loss did they dissect the audience, and the finding was blunt: the question was written in a way users couldn't answer, so "almost 75% of the users were not assigned any perimenopause phase." Those users, who had just clicked an ad promising to tell them their phase, got a screen that said sorry, we couldn't pinpoint it, carry on.
Flo rewrote the question with its medical team and changed the logic so that a user who still couldn't be placed moved quietly to the next section instead of hitting a dead end. "It was the third iteration based on the evidence, and it came back with 23% win." The team didn't trust it. A fourth, blackout test removed the new section entirely, and lost, which confirmed the uplift was real. Mike's summary: "All the iterations that were based on explanation loss, the one that was built on evidence One."
His rule for any failed test follows from that. "Explanation usually tells you what you want to be true," he says, whereas "the evidence tells you where to look." You iterate when you have new evidence, not a new explanation, because explanations are unlimited and chasing them is "this vicious cycle of, you know, experimenting."
Flo's marketing team found the need before product did
One of Flo's best onboarding branches began as an ad the product team hadn't planned for. The user acquisition team noticed that "a specific kind of narratives around enhancing your sex life, kind of started to gain traction at meta. And the numbers were pretty good, even with the existing funnel." Product took the signal apart and concluded there was a user need the onboarding had never explored. So they built for it: a dedicated branch that acknowledges which ad the user arrived from and asks a tailored set of questions so the narrative carries straight through. "It turned out to be a great revenue driver for us."
The mechanism matters more than the example. "Marketing team has a bad insight into things that resonate with your audience, and they can validate those insights much faster than you ever can do with your product funnel," Mike says (the transcript's "bad" is almost certainly "better"; check the audio). At Flo's scale, the marketing team can scan the market and test hypotheses in volumes a product funnel never could, so the practice is to share every change, weekly, whether or not it seems to matter: "you never know what insights will be valued by different teams." He doesn't claim the relationship is comfortable. "It's not like a happy marriage." It is, in his view, the only way to make the growth engine stronger than either team alone.
The Blinkist paywall didn't work at Flo, or anywhere Mike has tried it
Since joining Flo, Mike has been pitched silver bullets by product leaders at other companies. "We did this. We tracked 10% uplift in conversion rate. This is like your silver bullet. You got to do this. And we did exactly the same thing. Pixel to pixel, you know. And it just did not work." He has no tidy explanation. He does have a rule, which David restated as "borrow the idea, not the solution," and he thinks most product teams get it backwards: they "borrow the solution before understanding the problem."
The clearest case is the Blinkist-style paywall, the one that promises to remind you before your trial ends. It has worked for plenty of apps. "We did tested it at Luna, we did test it at the flower, and it did not work for both of the teams," Mike says, referring to Loóna, the sleep app where he was CPO, and to Flo. The trial toggle went the same way. His conclusion is not that the pattern is bad, but that a result from another product is not evidence it will work in yours: "You have to understand the underlying problem first."
In the full episode, Mike also explains why he thinks streaks have become a cargo cult, how Flo uses proxy events to make early calls on experiments when the real conversion is a delayed trial activation, and why putting partner mode, a feature with more than 2 million pairs using it, into the onboarding funnel was his biggest fail of the year.