I Made My Finance App Understand 14 Different Bank Formats — Here's How

My finance app DonFlow has no backend. Everything runs in the browser. That was a conscious choice — I wrote about why I chose IndexedDB over a server . But then came the hard part: users need to import their actual bank data. The Problem Nobody Talks About Every bank exports CSV differently. Column names, date formats, encoding, delimiters — nothing is standard. In Korea alone, there are 14 major card companies, and each one has its own export format. I could've said "use our template" and made users reformat their data manually. But that's the #1 reason people abandon finance apps — the data entry friction. My Approach: Universal Format Detection Instead of maintaining 14 separate parsers, I built a detection pipeline: Encoding detection — Korean banks love EUC-KR. The browser expects UTF-8. Step one: detect and convert. Delimiter sniffing — Some use commas, others tabs, some even pipes. Read the first 5 lines, count candidates, pick the winner. Header matching — Map known column pat