04 / LOCALIZATION · JAPAN MARKET
Localization is not translation: taking medical SaaS into Japan.
Entering Japan showed that trust is not built by translation alone. I turned what we learned on the ground — AI tone, kanji glyphs, brand familiarity, channels, and how to interview clinics across a language — into product rules and team methods.
01 THE CONTEXT
Translation is only the surface layer of localization.
A medical SaaS from Taiwan was entering Japanese clinics. Translation was the visible work; the design risk sat underneath it — in how clinics there judge professionalism, read text, and decide whether an overseas brand can speak for them to patients.
Learn how the market builds trust first; then decide what the product must change.
02 ASSUMPTION & TURNING POINT
The same kanji does not carry the same trust.
Japanese and Chinese both use Han characters, so we first read the differences on screen as a font-style issue and moved the translation and flows across.
Where it broke down
After launch, Japanese colleagues pointed out that some characters did not look Japanese. To a Japanese user, that reads as a foreign tool using the wrong glyphs.
What revealed the gap
Going through the interface character by character with them showed the problem was not how the font looked. Japanese kanji have their own strokes and conventions; we were using Chinese glyphs.
What I changed
I rechecked every Japan-market screen and made it a rule: Japanese-language fonts everywhere, with glyph shape checked in design review.
Localization is not only the content but the shape of the text. A rule has to be caught by a review step, or the next screen slips through.
THE TRADEOFFA review step that is easy to skip. A wrong glyph says “we have not understood how you read” before a user reads a word.
Explore the supporting design
03 THE SOLUTION
Keep tone flexible, and let clinics control it.
A friendly tone in Taiwan can feel too casual in Japanese healthcare, and emoji is the sharpest example: warmth in one market, a loss of rigor in the other. I did not bundle emoji into a reply style. It became its own switch, beside several tone options, so one product could fit both markets.
Tone under control
Emoji is its own switch and tone keeps several options, so each clinic decides how the AI speaks.
Glyphs by rule
Japanese-language fonts for every Japan-market screen, with glyph checks in design review.
Mascots for familiarity
Mascot characters, used by rule as part of the brand, so an overseas medical brand is easier to approach.
THE TRADEOFFOne more setting to explain. Locking a single “friendly” behavior into the product would have meant choosing one market over the other.
04 MAKING IT WORK
Turn what the field taught us into methods the team reuses.
On site, we learned that a full team in a small consultation room makes clinics answer carefully, and that translation slows follow-ups and smooths over hesitation. I changed the setup: fewer people in the room, roles agreed in advance, shorter and more concrete questions, time for the translator to confirm, and flow diagrams and demos to check understanding.
Fewer people hear the clinic first-hand. What we heard became more trustworthy.
- 01
Observe on site
In interviews, note the pauses, hesitation, and moments people check with each other — not only the answers.
- 02
Write it as rules
Tone, glyphs, brand, channels, and interview setup each became one rule the team can act on.
- 03
Fill gaps with tools
When sales needed fast support, build an internal tool with AI instead of waiting for development.
Explore the process and collaboration
05 IMPACT & OWNERSHIP
What changed through this work?
The findings became six rules the team reuses: an emoji toggle and tone options in AI Agent, Japanese typography guidelines, a mascot-based brand language, a CRM path from online awareness to LINE entry points to performance newsletters, a field-research setup for cross-language interviews, and an AI-built internal tool that let sales test and demo without waiting for development. The same rules supported onboarding the first 16 Japanese clinics with AI Agent.
My ownership: the localization judgments above and the rules that came out of them. Launch, sales, and translation were cross-functional; the 16 clinics are a product-level result reported through August 2026.
A CLOSER LOOK
A little more context.
Mascots as brand language
In Japan, banks, real estate, and medical services use mascot characters to make formal or unfamiliar information approachable. I made mascots part of the brand system rather than one-off decoration, with personality and usage rules, so landing pages, AI Agent ads, and DM materials feel familiar without feeling less professional.
One traceable path across channels
With the core team in Taiwan, online and offline had different jobs: websites, ads, and short videos to make AlleyPin / 1.Talk less unfamiliar; printed DM, fax, and in-clinic materials to add LINE entry points and QR codes; performance newsletters to show results and bring clinics back. Awareness → LINE entry → reactivation.
An internal tool built with AI
Sales needed quick answers to detailed product requests that did not belong on the roadmap. Rather than queue them for engineering, I used AI to build an internal tool sales could use to test and demo in market conversations — a usable bridge between product, engineering, and sales during early market development.
HIRING A SENIOR OR LEAD PRODUCT DESIGNER?
If this is the kind of work your team needs, let’s talk.
Based in Taipei (GMT+8), open to relocating for the right role. The form reaches me directly.