Hacker News Reader

Comment by hardtoreason | Hacker News Reader

I've used the ledger-likes extensively, and they have two major problems:

(1) One of the hard problems in personal finance is data capture and input validation. Programs like QuickBooks and Xero don't just hide the GL because they're being mean, they're doing it because even experienced bookkeepers make mistakes on the general ledger all the time. "Since it's a text file it's easy to read" solves exactly the wrong problem: you want it to be easy to write, and easy to write CORRECTLY.

(2) They're the sort of solution that seems fantastic if you don't actually operate at scale. But once you start dealing with tens of transactions per day it simply is not sustainable.

Safe to say: I'm not really a fan.

View on HN

Have you tried `ledger xact` command? This helps to create new transactions based on previous ones. I have a small wrapper[1] for this, namely "la", which enables me to do something like "la coles 5" to create a new transaction like previous supermarket transaction with the today as date, same from and to accounts, and $5 as amount. If I want to change something, it could be as complete as "la [YYYY.MM.DD/M.D/D] coles grocery 5 credit". If someone has tens of transactions per day it is very likely they are all similar and only little extra adjustment is needed.

Also I use a credit card for most of my expenses, and there are only a handful of bank transfers per month from my access account. Every first day of the month I pull out last month's credit card statement and log all the transactions. It usually takes less than 30 minutes.

[1]: https://github.com/roylez/dotfiles/blob/master/misc/bin/la

> One of the hard problems in personal finance is data capture

With hledger and vim, I typically search and copy a previous similar transaction, then search and copy similar items. I have a keybinding to change the date, so like 3\d changes the date of the transaction I copied to 3 days ago. Overall, being proficient in vim, capture takes very few keystrokes.

Since the files are plain text, I can easily e.g. list all products I've ever bought with a `grep | sort | uniq` pipeline, to check stuff like whether I put a product identifier under a wrong account or used different identifiers for the same product.

Also, since the files are plain text, I can track changes with git. That gives me clarity over changes in the sense that I can make a large change (e.g. reorganizing accounts) over historical transactions and review the changes before committing.

Regarding large changes, which I think is also a type of capture, I can easily make them with `sed -i ...`, vim macros, awk, etc.

> and input validation

After I've captured, running hb (which is a personal alias for `hledger balance`) results in hledger raising an error if a transaction is not balanced. I can also add an assertion `= X` to the end of account changes to assert what the account is supposed to hold after the change. The number comes from physically checking the account, whether that's by checking the bank app or opening my wallet and counting. hledger will raise an error when an assertion is wrong.

> Programs like QuickBooks and Xero

Being unfamiliar with QuickBooks and Xero, how would these programs do capture and validation better for me?

This is already after the data capture part. The hard part is to remember every little transaction I had until I have vim at my hand. The only solution seems to be working more or less is to take out my phone and type it as soon as the transaction happened. And I won't try to use vim on phone ever again..

I use GnuCash. For card transactions I simply import my bank statement .csv file every other week or so. For cash transaction, I use the app on my phone, otherwise I just forget how much I spent and where, by the time I get to my PC.

I never really got past the importing stage in GNUcash. I want to categorize the payments into different accounts, depending on how the money was spent. But there is no easy way of doing it. Or doing it iteratively.

When you import all your new transactions are presented. You can click (or ctrl-click to choose several) and match them to an account of your choosing.

It also memorises what transactions go into what category, so after you categorise, say "PURCHASE LIDL" as Expenses.Groceries, it will automatically match future transactions with the same name to groceries as well.

> The hard part is to remember every little transaction I had until I have vim at my hand.

Most of the time, there's no need to remember. Just put the receipt in your pocket, and when you get home you can further postpone it by stacking it with other receipts on a receipt spike. It's only the cash transactions that are best not left postponed for longer than a day or 2, but if you just scribble down the important bits on a scrap of paper, you can postpone it like any other receipt.

This is one of the reasons I use YNAB (for shared household budget) and a Google sheet (for personal budget).

I usually just hold onto the receipt for whatever I do until the end of the working day and get it entered that night. But if a receipt or my pay app isn't an option then I can enter the transaction right there on the phone before I forget.

Do you use the original or SaaS version of YNAB?

Currently on the SaaS version. It's much slower than I'd like on desktop web, truth be told, but it does the job and the mobile app is pretty good.

> and input validation

Emacs users should look into Flycheck integration, eg https://hledger.org/editors.html#flycheck-hledger . Real time indication of parse and balance errors as you edit is nice!

> hard problems in personal finance is data capture and input validation

It's also true for business accounting. Getting data into your accounting/ERP software is always a clusterfuck, and I've yet to see a process that didn't require many hours/days of reconciliation and posting manual journal entries to close a reporting period.

This is the data representation problem.

It's in the customers' best interest to have a universal format.

It's in no one else's (read: especially vendors) interest for there to be a universal format.

Which is why you generally only see it when it's mandated by a coordinating party in the middle (e.g. Medicare).

In finance, would this be the open banking standard?

It might be partly a US vs. UK/EUR thing, where the latter are more comfortable with compelling industry to just do a thing. And the former is very uncomfortable with that.

In the US, open banking would probably look more like "the IRS requires a bank to have accounts readable using these formats, or otherwise be able to provide same manually, when requested" to prod industry to do it.

Sort of.

The challenge is that no two businesses are really alike, so it's difficult to create a standard that can capture all the nuance of running a business. GAAP/IFRS-compliant accounting requires double-entry accounting, so many ERP systems share a lot of the same things (account, debit/credit, accruals, receipt of goods, month-end close) that are needed to spit out the three sheets and other artifacts for financial reporting.

A running joke in the industry is that you don't customize your ERP for your business, but customize your business for your ERP. You end up with a lot of "middleware" that essentially adapts your operational systems into the format needed by the ERP to do its job, and your staff will design workflow processes specifically centered around how the ERP works for things like closing the books or managing inventory.

http://xbrl.squarespace.com aims at this problem, I think ?

This is for exchanging reporting data, not all the things that lead up to that.

It can't tell you how to structure sales to account for taxes across multiple countries, or how to accrue for gift cards/liabilities, or recognizing revenue from annual subscriptions, or deprecating assets, or amortizing expenses. Each ERP and business will do this in slightly different ways.

But what it can tell you is how to represent those results when everything else is said-and-done.

Never had a problem with any of this. Yes, it is time-consuming, but so is everything in business. Maybe if I have someone to open a business in my name, do all the work, and I just sit and drink margaritas at the beach all day, none of work would be time-consuming for me.

But back to reality, everything in life is time-consuming. But, that is why it is so important to have a healthy profit margin and enough revenue in to hire a bookkeeper - so it takes you zero time.

But, even still, the real issue is that most business people don't know sh-t about accounting, don't know about accounting controls, don't know about so much, that they can't really evaluate anything anyways. How many companies have I been with or known someone that works at them that have been ripped off by embezzlement, because the company didn't know about controls. Too many. Rita Crundwell who was the Comptroller and Treasurer of the town of Dixon, Illinois is an especially juicy case of embezzlement. Crundwell stole nearly $2.5 million per year from the city. Crundwell with embezzling $53 million over the prior 22 years. And I think it was more, because she worked for a longer period than that, but no records were available. There was a great analysis of it, and wow, was it great. Just so fascinating.

I have a lot of experience in accounting and bookkeeping, so I really enjoy it. I don't know, it doesn't really take me a long time, I do it every day, so it takes me less than 4 or 5 minutes per day. Sometimes a little longer, sometimes less.

For data capture, what can help is just not aiming to record every cheque.

Record significant spendings, and account the rest as "other" - it should be the amount allowing to balance the ledger, given the current assets, income, and the significant spendings.

My problem with ledger was the absence of correct support for multi currency accounting.

Termux on the phone with custom ledger focused org capture templates has worked well for me

Now I just use a simple discord bot to capture, which is synced to a private git repository

Is your bot that does this open sourced? This seems very interesting

It would be nice to see how hard it would be to convert into a Matrix bot, at least the conversation would be end-to-end encrypted.

Sorry mate it's just an easy message parser built on discord.py, pretty bodgy code and no intention to OS

Messages look like

    &out soba shop
    Soba 800
    Beer 700

    &p #custom:category payee
    Item (int cost)
    Item (int cost)

Another data point: 95% of my data entry (all but cash) is automatic CSV download (using plaid2qif) and mostly-automatic deduplicating import (I tweak the import rules when needed). So, a daily `make import`. I have also run this from a nightly cron job. For most of us, that's a little too automatic; usually we want a bit more oversight and awareness.

100% this. Collecting and validating data should be so easy that you don't have to think about it. Otherwise I find I can't stick to any solution/tool for too long. If discipline is required, then it's too complicated.

This is why I've started looking into tools which automate the data collection. In the UK (and EU more broadly), after Open Banking got introduced there are many apps popping out which basically connect to your accounts (Credit Cards, bank accounts, etc.) and combine all data into a single place. Does Plaid not enable this in the US?

It's not "100% accurate" but most enable you to do auto-categorisation, smart reporting and budgeting as well. They are not open-source but do much more than I'd ever be able to do with PTA (not because it's not supported but because it requires effort). Of the 2 bad solutions, they seem to be the less bad for me.

Some example apps: Emma, Yolt, Lumio.

1. PTA in itself does not provide you all the tools to validate these, but it's low-level enough that you can implement any validation you want on top easily, if you're proficient with CLI tools. I don't think I would use it at the enterprise level simply because they're less commonly used, but for personal finances I find it sensible enough.

Which mistake do you think is more common with PTA than with other programs?

2. What exactly would not be sustainable with PTA which would be with other programs?

Yes, it's much easier to just have accounting data in a relational database. It's much easier to work with (edits and arbitrary analysis/reporting), rather than having to use/learn some bespoke tooling and data format. It's much easier to associate entries with metadata or categories, and much simpler to add editing GUI on top if you want to, or integrate with third party data sources.

(I wrote a text based ledger tool myself, and then switched to storing data in PostgreSQL)

I dunno, I kinda disagree: I am not writing transactions manually, they are written by my brokerage parser, bank statement parser, blockchain parser, etc.

The reporting and analytics functionality of the ledger-likes is AFAICT unrivaled, and is the REAL reason to use them (not the ostensibly easy-to-use “plain text” general ledger aspect, about which I agree with you)

The reason I use Xero is the integration with my bank, credit cards, invoicing, and payroll system.

> Programs like QuickBooks and Xero don't just hide the GL because they're being mean

What is the GL?

General ledger

https://en.wikipedia.org/wiki/General_ledger

Reply on news.ycombinator.com