BoldVoice

Charging from launch reveals who the real customer is. BoldVoice launched in the first weeks of YC with $10 a month and a one-week trial, against the 2021 grow-first mood. Willingness to pay correlates with the problem resonating, so paying users showed the ICP to focus on.

It keeps the team from chasing feedback from people who will never be the customer. Anada Lakra sees a one-week trial as a stronger signal than a fully ungated product. "Charging for the product is a way to already figure out who that ICP is."

Sub Club▶ 13:53

Launch with a plain paywall and put effort into onboarding and product. The launch offer was a standard $10 a month plan with a one-week trial, so the team built onboarding and the core product instead of running pricing tests. Paywall and pricing variants came later.

A high trial cancellation rate means people found no value inside, and two more onboarding pages won't fix that. "Do optimize the paywall. It's important, but don't overoptimize the paywall compared to the actual product value."

Sub Club▶ 14:50

Duolingo's freemium playbook does not fit a high-intent audience. Duolingo gives away basics of every language to a huge free base and monetizes a slice. BoldVoice ignores beginners and serves intermediate to advanced professionals whose alternative is one-on-one coaching costing hundreds of dollars an hour, which supports a $150 to $200 a year price.

"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."

Sub Club▶ 29:05

Working professionals beat college students as the target audience. BoldVoice tried the campus playbook for international students, but they improve faster on their own among native speakers, have less ability to pay, and haven't yet felt the career cost of their accent.

"The audience that resonates the best with this were people who were in the workplace already."

Sub Club▶ 34:10

A free first day works as a reverse trial behind a paywall. If a user dismisses the paywall, they get the speech assessment results, a practice set and a coach video free on day one; the 7-day trial is required from day two.

This suits skeptical users who dislike entering a credit card while keeping the trial paywall in place.

Sub Club▶ 24:40

Removing the paywall's exit lifted trial starts but lost money to refunds. One test removed the X button; another let users dismiss the paywall, then showed an X% off offer they could not close. Both raised trial starts because users had invested in onboarding, but refunds spiked by more than the gain, making both clear losers.

Forced starters asked for refunds angry, building negative sentiment. Exact refund numbers were not given. "Anything that kind of forces or tricks people into starting a trial, you'll pay it when it comes to refunds."

Sub Club▶ 19:40

Judge tests on net revenue after refunds per exposed user, on matured cohorts. BoldVoice waits one week for trials to convert plus at least two more weeks for refunds to settle before picking a winner. Trial start rate became a default scoreboard only because it arrives first and reaches significance fastest.

Refunds come only after the charge hits a bank account. The cost is speed, not discipline: sitting on a result for a month delays learnings, but Anada has not had to reverse a shipped decision. RevenueCat's lookback window of over a year makes checking back easier.

Sub Club▶ 23:40

Defaulting to annual plans shows channel payback right after the trial. The paywall now leads with an annual plan of about $150, a monthly fallback, and a $200 a year tier with AI chat. Annual cash arrives after the 7-day trial, so per-channel spend and subscribers show LTV over CAC for the first year at once.

Monthly plans leave it unknown whether a subscriber stays one month or 24. The data showed TikTok produced flimsier, lower-loyalty subscribers than other channels.

Sub Club▶ 50:10

A free, ungated accent-guessing tool funnels people into the app. BoldVoice trained a model that guesses where you're from in 10 seconds of speech and shipped it as a separate web app at aoracle.com with no download, gate or trial. After an angel investor posted it on Threads, about 100,000 people in Korea tried it the next day.

Results show a snippet of what the full app offers, which raises conversion into the app. Adding city-level results gave another viral boost.

Sub Club▶ 44:40

Humorous viral content brings spikes but lower-intent users. A TikTok trend of people crying at low app scores spread without ad boosting; one creator reached about 300 million views in a few months.

Viewers had less intent than paid-ad or word-of-mouth users, so BoldVoice treats humor as a companion to paid marketing.

Sub Club▶ 47:30

In-house models on proprietary data keep general LLMs from copying the product. An in-house audio ML team trains models from scratch on BoldVoice data labeled and transcribed by expert speech coaches. ChatGPT and Claude still cannot give the same feedback.

The bet: non-native English speech is too niche for general models yet a huge market for one company. The idea came from 2021 audio AI, before ChatGPT, making coach-precise feedback look possible.

Sub Club▶ 53:50

Expand in concentric rings from the sharpest pain point. BoldVoice started with accent training, the spikiest pain, then after a Series A added fluency, confidence, idioms and grammar.

It declined to serve native speakers, whose need is more of a one-off nice-to-have.

Sub Club▶ 56:00

Every employee does at least two 15-minute user interviews a week. Automated invites go to a varied sample, from three-year subscribers to new trial starters, and book whichever teammate has calendar space. Founders, engineers and growth staff all take calls.

Sub Club▶ 37:00

Layer research methods so happy users don't dominate feedback. Random-sample calls sit alongside a weekly super-user group where engineers demo features, in-office sessions watching people use the app, a 3-question email survey and a cancel-flow question.

"You're always going to get more feedback and you're going to get more people to jump on calls who are happy users."

Sub Club▶ 41:40

Founder-type owners, one weekly meeting and AI tools keep the team small. BoldVoice hires doers who need little direction and holds one company meeting on Mondays. Everyone uses Claude: the CEO ships pull requests and the designer's prototypes arrive about 80% ready for engineers.

The $10M ARR and profitability claims come from the founder without supporting figures.

Sub Club▶ 50:50

A concierge pilot and a six-month deadline proved the idea before code. Six months before graduating from Harvard Business School, Anada Lakra turned down a McKinsey return offer and gave himself until then to get a term sheet. A speech coach recorded videos for about a dozen beta users, who emailed recordings back for manual feedback; a roughly $500K term sheet followed and sped up YC.

An earlier recruiting startup built before talking to users taught him to involve customers early. "We're just doing the dry run of super manually without any lines of code being written."

Sub Club▶ 11:40