Replika · B2C · AI companion chatbot
Fully translate and adapt an app for Japan with a Japanese speaker
Didn't work at all; harder than expected
Failedlocalizationmarket-selection
What they did
Made a big bet on Japan with a Japanese-speaking team member doing full localization and culturalization of the app.
What happened
Didn't work 'in any way'; harder than expected.
- Stage
- scale
In their words
Sub Club by RevenueCat · Make Ugly Ads to Grow Your App – Yuliya Lennox (Solid Starts)“I can say that in Replika we did make a big bet on Japan. We even have person that knew Japanese in the team and we did the actual localization, the culturalization of the app didn't work in any way because it is harder than we think.”
Related topics
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
- BetterMe: Keep testing new ads until one sharply cuts costs, even if current ads work
- BetterMe: Run ugly ads that show the product and the pain in the first secondOne ran for years; called 'the best ad ever'; no numbers
- Solid Starts: Run a trendy AI 'fruit babies' ad for a doctor-backed baby feeding brandBacklash: parents questioned whether to trust the brand
- EWA: Learn English: Have the founder join every marketing meeting and write the ads himselfApp reached #2 after Duolingo; link to this is loose
- EWA: Learn English: Win your home market first, then add languages one by oneLiked locally, then expanded; later #2 after Duolingo
- Test new ads at $10/day in Indonesia, not $1,000/day in the USSame test signal at ~1% of the cost
- EWA: Learn English: Run paid ads from the start alongside organic growthOrganic-only apps plateau; paid added late often ate into organic
- Palta: Share test results across apps and with a weekly peer group
- Make the paywall's continue button blinkRaised conversion, secondhand; no numbers
- Ladder: Sign a celebrity like Hilary Duff for brand adsBig influx almost at once, secondhand; ROI hard to measure
- Solid Starts: Hire people obsessed with the problemCredited for product quality; no numbers
- Higgsfield: Have the core team work ~17-hour days, six days a weekSaid to lead the field now; no numbers
- Translate the app into many languages with AI and test nearby marketsHer biggest win of the year; Turkey converts well; no numbers
- Sell a $1 plan after a 60-question quiz, then stack add-ons to ~$350Buyer got half back after demanding a refund; invites regulation
- Run three apps at once with one teamBiggest fail of the year: attention spread too thin