One Number I Trust: Plain-Text Accounting for a Multi-Currency Household

Summary

Two people. Eighteen accounts spanning checking, savings, credit cards, investments. Three currencies. Twenty minutes of work every week.One net worth number I actually trust.The payoff: A single, trustworthy net worth number growing over time.No app did exactly what I needed, so I built my own personal finance system using plain-text accounting principles and a powerful Python library called Beancount. This post shows you how I handle imports, investments, multi-currency, and a two-person view.How I got hereIt all started during the 2021 tax season. I had blocked out an entire weekend and was juggling statements, trying to compute capital gains, stressing about getting the numbers mixed up. “This is chaos”, I thought. “There must be a way to simplify this with automation”. Being a software engineer, I did what felt natural and hacked together a bunch of scripts on top of a database.Though it worked and I kept using it day-to-day, by the next tax season the cracks became obvious. The code was hard to debug, random transactions went missing, and worst of all, the balances the scripts computed didn’t match the balances on my statements. I tried to fix it but the more I tried, the more I felt lost about what the system was really doing. Eventually I just gave up.Why did I fail so spectacularly? My entire approach was flawed from the start! I’d ignored centuries of accounting wisdom and repeated fundamental mistakes humanity solved long ago. So I learned from my mistakes and did the research. And over time I incrementally discovered double-entry bookkeeping, plain-text accounting and Beancount.Fast forward to today, and I have a flexible, powerful, and private system, fully customized to how my brain works. Most transactions import automatically from PDF statements (counterintuitively, it’s often more reliable than CSV!). Tax time is a simple matter of checking always-fresh reports and copying numbers over. The weekly ritual is simple: download statements, categorize tr...