The target user was a factory worker in Chennai on a $60 Android with 2G data. Every React Native default assumption breaks against that user.
At Unipe (Payrocket) I built the React Native app for earned wage access used by blue-collar workers in India. Salary disbursement, KYC, financial workflows, all in the pocket of users whose phones were three generations behind, whose internet was intermittent, and whose primary language was not English. Here is what I learned.
Metro's default bundle was 4.2MB before we shipped a screen. On a $60 Android with 1GB RAM, that alone caused startup jank. Aggressive tree-shaking, lazy loading non-critical screens, and switching to Hermes cut startup time in half. This was the single biggest UX win we got, and it did not touch a feature.
Our first KYC screen was designed like every other Indian fintech at the time. Upload PAN. Upload Aadhaar. Take a selfie. Enter details. It dropped 70 percent of users at the first upload step.
The reason was invisible until we watched a user in person. Their phone camera resolution was so low that the OCR could not read their PAN card. They tried three times. On the fourth attempt they closed the app. We fixed it by falling back to manual entry after two OCR failures, with a mode that guided them through the fields one at a time in Hindi. Drop-off went from 70 to 22 percent.
Every screen had to work in Hindi, English, Tamil, and Telugu. What nobody told me was that Hindi is 30 to 60 percent longer than English for the same message. Any button, any input label, any error copy designed for English broke visually in Hindi.
A user in this segment has probably been scammed by a fake loan app in the last year. They do not trust colored buttons that say "Withdraw Now." What they do trust is a plain sentence that says exactly what will happen next, in their language, before they tap anything.
When your user has been burned before, the design job stops being about delight and starts being about respect.
This is the single biggest gap between how frontend engineers in tier-1 cities design and how the actual users in India need apps to work. If you get one thing right for this audience, get this.
Why I Am Building AI Consumer Products Solo in 2026
For the first time in software history, one engineer can ship a real consumer AI product in a weekend. Here is why I left founding engineer roles to focus on it.
How I Shipped an AI Product in 48 Hours
The exact process I used to take SoulCompass from idea to live AI product in a weekend. What I cut, what I kept, and what actually matters.