BoldVoice：停止过度优化Paywall，全员每周2次用户访谈

BoldVoice tested two changes that every paywall optimizer has been tempted by. The first was a hard paywall: remove the dismiss button so nobody can proceed without starting a trial. The second was a softer trap: leave the X on the main paywall, but if a user dismisses it, show a percentage-off offer they can't close.

Watch on YouTube

Both looked like wins. "Both of those look so promising on the trial numbers," says Anada Lakra, BoldVoice's co-founder and CEO. "And it makes sense, right? People have already put in all the time in the onboarding flow. They actually want to see what this app is all about." The team's instinct was to ship both to everyone.

Then the refunds arrived. "The refund spiked by way more than the gain in start trials," Anada says. People who feel forced into a trial ask for their money back the moment the charge hits their bank account — and they leave with a grudge. "Anything that kind of forces or tricks people into starting a trial, you'll pay it when it comes to refunds."

The fix was a different scoreboard. BoldVoice now judges paywall tests on net revenue after refunds, per exposed user, on a matured cohort — which means waiting at least an extra week for trials to convert and two more for refunds to settle before declaring anything. It's slower, and Anada admits the hardest part isn't discipline but patience: sitting on exciting data for a month or more. Asked at the end of the episode for her worst change of the past year, she doesn't hesitate: "The paywall stuff. Just stop over-optimizing every single word and every button and everything. Optimize them for a little bit and then move on to the actual product."

Every employee does two user interviews a week — including the engineers

BoldVoice's user research isn't a function. It's a calendar rule.

Automated invites go out to a deliberate mix of users — subscribers of two or three years, people who started a trial last week and converted, people who didn't — asking for a 15-minute call. The invite automatically books whichever team members have space in their calendars. "Every single person on the team — myself, my co-founder, our engineers, our growth people — everyone is in at least two user interviews a week," Anada says.

Two more layers sit on top. A super-user group meets roughly weekly, and engineers demo unshipped features to them, so heavy users shape the product before release. Users also come into the office so the team can watch what they tap and where they linger. "Being able to actually verbally talk to users and actually see them interact with the product, I think is an underrated skill."

Anada is candid about the bias in this setup. Happy users book the calls. "People who did not have a good experience are going to be less likely to jump on those calls." For them, BoldVoice uses a three-question survey answerable by replying to an email, and a short flow at cancellation asking what went wrong. Without that correction, she says, you end up making happy people happier and building a product that gets narrower and more technical instead of wider.

Why BoldVoice charged from day one — and won't copy Duolingo

BoldVoice launched in the first weeks of Y Combinator's summer 2021 batch and charged from the start: $10 a month, one-week free trial, no experimentation. That was the point. "Charging for the product is a way to already figure out who that ICP is," Anada says. Willingness to pay correlates with the problem resonating, and it kept the team from chasing feedback from people who would never be customers. The energy went into onboarding and the core product instead of the paywall.

Pricing has since matured — BoldVoice leads with an annual plan at around $150 a year, offers monthly as a fallback, and has a $200-a-year tier with extra features like chat — but the philosophy hasn't moved toward freemium, and Anada is precise about why. "The worst thing that someone can do is take some playbooks and some elements of a playbook from one app, combine it with a playbook of a different app, and create this mishmash strategy."

Duolingo's freemium base works because it teaches the basics of every language to a huge audience of hobby learners, and monetizing a fraction is enough. BoldVoice intentionally ignores that audience. Its users are intermediate-to-advanced professionals for whom the alternative is expensive one-on-one coaching with an accent coach: "totally different ICP, different value proposition, different willingness to pay." The same logic ruled out the cliché college-campus launch, which BoldVoice tried. Students had less money, less urgency, and hadn't yet sat in the meeting where a colleague got listened to instead of them.

What BoldVoice has instead is a narrow free taste: dismiss the paywall and you get the speech assessment, the first daily plan, and a coach video, then a seven-day trial to continue on day two. David Barnard calls it a reverse trial. Anada calls it a middle ground that doesn't cost you the onboarding paywall for the people who are already ready to pay.

One Threads post, 100,000 people in Korea

BoldVoice's in-house ML team, working with the Hollywood accent coaches who help label its training data, built a tool that guesses where you're from after about ten seconds of speech. They shipped it as an ungated web app — no download, no trial — called the Accent Oracle, and Anada mentioned it in her monthly investor update.

"One of them was an angel investor. It was Korean, and he was like, 'Oh, this is cool. I'll put it on Threads,'" she says. "And then I believe the next day we had 100,000 people in Korea try this thing."

The lesson she draws isn't "build a viral tool." It's that you can't engineer virality, but you can add levers to something that already has it: a city-level guess re-spiked the Oracle, and showing users a preview of the fuller results waiting in the app lifted the number who made the jump. A TikTok trend of users crying at low scores gave the brand another unpaid spike — hundreds of millions of views for one creator, by her estimate — with the caveat that humor brings reach, not intent.

$10M ARR, 10 people, one meeting a week

Anada’s explanation for roughly $1M of ARR per employee is unglamorous. Hire founder-type doers who don't need direction. Hold one company-wide meeting, on Mondays, and leave the rest of the week for building and talking to users. Default users to the annual plan so paid acquisition pays back within the trial window and every channel's return is legible the same week.

And let everyone ship. "Everyone in the company is using Claude. For the first time in my life, I'm actually shipping PRs myself, which is crazy. I'm not a technical person." The designer takes Figma work through to prototypes that are about 80% of the way to production before engineers touch them.

The part that keeps the small team defensible is the part that took longest to build. "Still to this day, you cannot talk to ChatGPT or Claude or whatever have you and get this kind of feedback that BoldVoice gives you," Anada says — a deliberate bet that non-native English pronunciation is too narrow a niche for general models to chase and far too large a market to ignore.

In the full episode, Anada also explains why BoldVoice is expanding from accent into the "penultimate mile" of spoken fluency, why she considers "accent" an almost anti-viral word, and what a self-imposed six-month term-sheet deadline did for her focus.

Guest links

• Anada Lakra on LinkedIn
• BoldVoice on LinkedIn
• BoldVoice