BoldVoice · B2C · accent coaching app
Build a free web tool that guesses where you're from after 10 seconds of speech
~100K people in Korea tried it the day after one investor posted it
What they did
What happened
- Stage
- scale
In their words
Why BoldVoice Charged From Day 1 and Ignored the Duolingo Freemium Playbook — Anada Lakra
David BarnardSuch great advice. Gosh, there's so many things I want to talk about. We don't have a ton of time left, but maybe we can speed run some of the virality and growth things y'all have done because you've gotten some really cool growth wins and some viral moments and things like that. What would you say were the maybe top three most impactful lessons on the growth side or top three viral moments?
Anada LakraWell, for sure the top one was the Accent Oracle that we built. Our team had this idea that if we could build a tool that could predict where a person was from just based on 10 seconds of them speaking, it could have this kind of viral potential. And so we built and trained a model in-house to do exactly that. So I think this also speaks to ... Again, a really important part of our success is that we have an audio ML team in-house building these and training these really differentiated models from scratch with our own data. And we also have the expert speech coaches who are a huge part of how we label, transcribe and just train these models to be better. So that's a really big differentiation in our product. And the Accent Oracle was built based on one of those in-house created models, and it worked. It was the most magical thing I think we've done.
Anada LakraAnd the best part about it is that the way that it spread is that ... Every month I write the investor updates. I've been doing it for five years. That month that the Accent Oracle was built, I sent it out. I think even the day after or so, to our investors. One of them was an angel investor, it was Korean, and he was like, "Oh, this is cool. I'll put in on Threads." And he posted it on Threads, and then I believe the next day we had a 100,000 people in Korea try this thing. It's again, wild and unlike anything we've seen before.
Anada LakraSo of course, when that happened, we're like, "Let's make this better. Let's invest a lot of time in making the model more accurate, making sure it doesn't crash." And all these things. And it was a really wonderful way to get people familiar with the product, and we made it ... There was no gate, there was no free trial. It was a pretty magical experience. You don't even have to download the app, by the way. My co-founder actually made this decision to just put this in a separate web app, where you just go to accentoracle.com. You just record your voice, instantly it spits out your feedback or your results.
Anada LakraSo it was optimized for this frictionless viral moment that then exposes people to BoldVoice as a brand and then invites them into the app to continue discovering more things about their speech and then actually improving and working on their pronunciation and speech as well.
Get tactics like this every Monday
New tactics from the week's founder interviews, each linked to where it was said.
More from this episode
- Charge $10 a month with a 7-day free trial from launch dayFounder says it filtered out non-buyers and freed time for the product; no numbers
- Judge paywall tests on revenue after refunds, waiting 3+ weeks past trial startCaught two trial-start 'wins' that lost money once refunds came in
- Remove the close button so users must start a trial to continueTrial starts rose, but refunds rose by more; net loss
- When users close the paywall, show a discount trial they can't closeTrial starts rose, but refunds rose by more; net loss
- Let people who close the paywall use day one free, then ask for a trialFounder calls it a good middle ground; no numbers
- Show the ~$150 annual plan first so ad spend pays back when trials endCash lands 7 days after install, so each channel's ROI is known at once; no numbers
- Market to international college students firstWeak fit: lower willingness to pay and less urgency; moved to professionals
- Sell to working professionals with good English, and skip beginners entirelyFounder credits this focus for ~$10M ARR; no split by segment
- Auto-invite users to 15-minute calls booked with whoever on the team is freeEvery employee, engineers included, does 2+ user calls a week
- Have engineers demo unreleased features to power users on weekly callsFounder says it validates complex features before shipping; no numbers
- Ask unhappy users three questions by email they can answer by replyingOffsets the happy-user bias in interviews; no numbers
- Add city-level results to the free accent tool to spark a new viral waveProduced another viral boost; no numbers
- Show a sample of the full in-app results to push web users to the appFounder says it boosts web-to-app conversion; no numbers
- Ride a TikTok trend of people crying over their low app scoresOne creator got ~300M views in a few months, no ads; lower intent than paid
- Grow from accent training into overall spoken-English fluency after raising a Series A
- Hire self-driven founder types and hold just one team meeting a week, on Mondays~$10M ARR with 10 people; founder credits hiring first
- Get everyone, even the non-technical CEO, shipping code with ClaudeRoughly doubled team productivity, per the founder; her biggest win of the year
- Recruit a Hollywood accent coach and run a manual email pilot with 12 usersLed to a ~$500K term sheet and YC; the coach still works with them
- Show a countdown of free scans left before the paywall appears
Tags: web, threads, free-tool, viral-tool, accent-oracle, no-download, investor-update, in-house-ml