Hacker News new | past | comments | ask | show | jobs | submit login

How are people automating the data import? I can't imagine someone entering everything by hand. Lots of places don't even provide an export file you can work with... most of them offer a PDF.

It seems like most financial places rely on Plaid for the data integration, but that's a paid service I don't think Open-Source or free personal finance apps would use.




Some years ago I built a custom tool similar to this which downloaded financial data from my Chase bank account and converted it to beancount (another text-based accounting tool).

Chase charged like $10/mo or something like that for using OFX to download your bank statements (which is pretty ridiculous considering what it is). Eventually I abandoned it because none of the other bank accounts I needed to track offered OFX or anything similar that I could find, and just gave up.


As far as standard checking/savings/credit card accounts go, I haven't had one that didn't have a csv export option yet. Some of them are very buried and much harder to find and use than I wish, but they are there at least with the banks I use.


Most banks/CC have some sort of export - be it in CSV, OFX, Quicken, etc. Then you import it using your financial story.

The real pain is in the categorization (was this Groceries, Supplies, Dining, Medical Expense)?

I use KMyMoney which usually picks the same category as the last transaction from the same place. Saves some of the work, but it's still painful. I then wrote a script to export from KMyMoney to ledger format.


Reckon works well enough for me.

https://github.com/cantino/reckon


Categorization seems like low-hanging fruit given the age of AI we are in today.


> Categorization seems like low-hanging fruit given the age of AI we are in today.

AI can assist, but that's it. When I go to a Kroger's store, I may buy food. Or I buy supplies. Or medicine. Or gifts. Or something else. And almost always a combination of these. We are already at 5 categories, without mentioning others. All my CC will report is a transaction at Kroger's. How is an AI supposed to guess how much I spent on each?


Take a picture of the receipt and a picture of your groceries, pipe them to appropriate models?


You think so? They have such inconsistent reporting, and the data is the exactly hard kind to aggregate massive sets of.


I enter everything manually. I live in SE Asia where there's no other option, really. For some of my banks there's not even a CSV/Excel download option, just a PDF "balance statement".

I (mostly) don't find it too onerous. But it is very much Inbox Zero. If you keep on top of it then it isn't much additional effort. While I'm waiting for the cashier to make change or print a receipt or while I'm walking out of the shop I can enter the transaction. But if I let a few days of them build up it becomes a more annoying 5-10 minute effort.

Also you need to do manual entry for anything you pay with cash anyway.

But the manual approach starts to fail in a family unless all involved are on-board with the data entry. (Which is almost never the case in my experience.) In practice that means all of my wife's expenses are a bit of a black hole.


If the PDFs have data in table format, you can use image to table generator of nanonets or Google or Microsoft Excel app to convert it into csv and enter data.


I use hledger and I enter everything by hand.

I used to dread it too. But actually, in normal life, the volume of daily transactions is not that large. Besides, filling out a journal disciplines you. And VS Code does autocomplete for accounts and payees.


Can I ask how you get the autocomplete in VS Code? I've never managed to find an extension that does this, and by default the autocomplete only works on previously seen words (so it gets split by the colons in account names)


I did nothing. It just works out of the box for me.

But. I keep my journal files in a folder and open it using the VS Code project file.

File's name is ledger.code-workspace

{ "folders": [{ "path": "/path/to/journal/directory" }] }


> only works on previously seen words

Yes, that's just it. I do not use anything like a language server.


You can sign up for a developer account and have up to 100 accounts that you access.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: