Comment by WalterBright | Hacker News Reader

Comment by WalterBright | Hacker News Reader

Doing my taxes this year, I find it amazing how incompetent most financial software is (including Quicken).

For example, take a simple task: download all your financial data for the last year. Log into your bank's site. There is no button for "last year". You have to laboriously enter 01/01/2015..12/31/2015. Then you're given a choice of formats. Download ALL of them, because sure as shootin', Quicken will only import one of them, and that's one is different for every different account! (Stunningly stoopid.)

Then try to get the statements, just try, my pretty. There is no "get me all the statements from last year." Nope. You download them one at a time, and some of them download to the same file name! So it's clicking around, then rename the file to "January 2015", then repeat for the next one, etc.

Oh, you wanted the tax 1099 forms? That's a whole 'nuther section on the web site.

Buy anything from Paypal? This is the killer. Paypal lists the payee as "Bob". Your credit card statement lists the payee for the same transaction as "Carol". The credit card downloaded data file lists the payee as "Ted". And, Quicken somehow calls it "Alice". Can you believe it in 2016? Naturally, you cannot download the full information about the transaction anyway. Download the CSV file, and Paypal randomly leaves out information you have to manually edit back in, like the item description which appears on the web site but not in the CSV data.

And why does my bank list the payee for a check as "Check 236"? The bank knows who the money got sent too. It's all on a computer.

What gets me is when I see software designed this way and then start imagining how much money was spent developing it, and fellow software architects sitting back with their arms folded staring at the screen thinking what a good job they did.

sitting back with their arms folded staring at the screen thinking what a good job they did

Isn't that a basic job requirement for a software architect at BigCo?

Yes they used several UML diagrams, several design patterns. They couldn't code a line of it but the drawings look nice

This is so close to reality it hurts

Every year I call up one or two of their tech support numbers, and suggest they add a button labeled "download everything I need to give to my accountant for the last tax year." Every year they are astonished at this suggestion, say what a great idea it is, and next year nothing has improved.

Am I really such a unique snowflake that nobody else even thinks of this?

Maybe I should patent it. It's obviously not obvious!

There's no incentive, but you should probably really be pushing the Accountants of (Country) professional org (or whatever it's called) to support this.

They could register a branding for compatible download systems.

If we built that software, how far would we be from not needing accountants?

Collecting the data can be automated. Classifying it, not so. I.e. what is a business expense, what is not, etc., all require knowledge that is not reasonably available to the computer until a human classifies it.

For example, I fill up at a gas station. Is that for a business purpose, or not? The computer cannot know. An awful lot of such decisions need to be made.

The thing is different people have different tax needs. Some people have very tax filings while others are very complex and some things can't be downloaded at all. It would be great but it's just not possible. FYI I work in a related field and I can tell you some tax filings are complicated!

Also not every source generates valid data files :(

Just because it's difficult to handle 100% of use cases doesn't mean you can't easily solved 80% of use cases. The suggestion the parent and grandparent comments are making is simply to add a button which combines the previous year's statements (understanding that not everyone's fiscal year matches the calendar year) with all the documents that are already in a separate section named "tax documents" into a single download. More generally, they're just suggesting things like naming the file for January's statement as e.g. "2016-january-statement.pdf" instead of the same generic name as every other month's statements.

Saying that those suggestions aren't valuable or worthwhile just because 100% of use cases can't be accounted for probably isn't the best way of deciding which features should be implemented (and if it were, then it seems like most of their current features should have never been built since they rarely seem to handle even 80% of use cases).

My brokerage account allows me to search for statements and then check off the ones I want to generate a PDF for, they then get combined into a single PDF. Of the three banks I deal with it was the only one with this feature made catching up on my statements so much easier.

It certainly is possible because every banking site has a fixed set of downloads for a particular year. There's no problem at all putting them in a zip file and making that available for download.

For example, for a checking account there are:

1. The monthly statements 2. The transaction data in CSV, QIF, QXF formats 3. The tax filings (1099) 4. Overdraft notices, safe box bills

"staring at the screen thinking what a good job they did."

Do you work in programming in anything other than greenfield projects?

I'm sure they weren't nodding in smug pride, but weeping and drinking themselves into a stupor many nights, and sometimes during the workday. They know. At least, the better of them know.

I feel like those are the codebases that still cause you to wake up in a cold, guilty sweat 5 years after last modifying it.

I remember our bank changing our account type (Number stayed the same, they just changed the fee structure) so they expunged all our transaction history from the site, we could only request paper copies be sent out.

I ended up scanning them all, writing a script that split the images up and OCR'd them, then formatted them correctly into and importable format in a single file. It turned out to be much faster and easier than the CSV data provided by the bank (Actually having transaction descriptions on your imports helps!). So much so that for the next couple of years instead of getting the CSVs from the bank, I got PDF copies of our paper statements and fed them into the same script.

Wow, that's a quotable story, and says a lot the status of our field as a whole.

If only there was a web plugin or some other web client software that could login to your bank account for you and fetch everything you want, in convenient formats, with nice labels. It would support the weird interfaces of thousands of different banks (and PayPal)...

I'm not sure if you are being coy but didnt Zenbooks get sued or attacked by some banks for doing exactly that?

Plaid basically does just this, though as an API. And it doesn't support thousands of banks but just a few major ones.

I've been using Mint for this purpose and so far it almost meets my expectation. The automatic categorizing works well.

Quicken attempts automatic categorizing, but botches it so badly I have to fix 90% of them manually.

GNUCash supports QFX and OFX, but banks seem unprepared to actually support an API endpoint like that. End result: Mint and whatever systems my retirement funds started using to 'plan your retirement' ask for username and passwords to other accounts. A rare few banks allow for OAuth style read-only tokens, but it's not like my 403b is selected based on security principles.

Just wait until the payee has a comma or a quote in the name and you download a CSV and the programmers can't even properly format a CSV and it's broken in Excel.

Makes you wonder what part of their system might be susceptible to a payee name like lil Bobby Tables.

For those who didn't get the reference:

https://xkcd.com/327/

And banks like BofA only provide 13 months of statements so if you do your taxes in March, you're out of luck.

The $99 drive sitting on my desk could store the complete records of BofA going back to the beginning twice over (without compression).

I just don't get it.

That's especially aggravating because the IRS can audit your return going back 3 years. And you're right, it's a trivial amount of data to store.

I had a lot of trouble with Paypal because it kept insisting that the transaction data I requested was "too large" for it to handle. It was about 40k of data, less than the web page it served telling me that.

Oh, they store it, you just have to pay extra for a complete record. In a more perfect world, this is because it takes effort to retrieve your history from offline storage.

Reply on news.ycombinator.com