undefined | Better HN

0 pointsgamblor9565y ago0 comments

From first principles we should be able to form a generalist approach to solving sales tax. Should be an algorithm in the public domain as far as I'm concerned. : )

Reasoning from first principles only works if you're Elon Musk and you don't actually have to go beyond a superficial understanding of the problem domain. Unfortunately, there is no "generalist approach" to sales tax or VAT. Every jurisdiction has different rules and rates. The U.S. has more than 3000 different sales tax jurisdictions, of which more than 1000 levy sales tax on digital goods and services even on out-of-state sellers.

And that doesn't take into account the dozens of different definitions of what is a "digital good." Here's a surprise for the unwary: in more than a dozen states, downloaded software is treated as a physical good.

I'd take a solution that got me 80% there.

In this case, the last 20% will take 99% of the time and effort. And then you'll have to make sure to do that work all over again next year, and every year, when the sales tax/VAT laws are updated.

0 comments

yarcob5y ago

Also, noone will notify you when a sales tax or VAT law changes -- you have to find it out on your own :)

It's not an impossible problem, it's definitely doable, but I'm pretty sure you need a couple of people and won't be able to do it all on your own.

If you're just a small seller, you can "fly under the radar" and just handle taxes in your own jurisdiction.

But as you get bigger, you need to handle every place that you sell to. Especially if you resell your solution to others.

foxhop5y ago

I think a general system could be designed such that most of the classifying work is done up front and also shop owners could opt-in or add additional rules to the algorithm. I think the 20% could be squeezed and solved a number of different ways especially in the name of the public.

gamblor956OP5y ago

Yes, a "general" system can be designed. In fact, they already exist (see, for example Avalara, which does exactly what you aim to do). It took hundreds of developers and hundreds of accountants and lawyers to make these systems work, and the companies that run these systems make thousands of dollars a year from each business customer. And businesses that sell to multiple states/countries gladly pay, because it would cost even more to try and build such a system in-house.

I think a general system could be designed such that most of the classifying work is done up front

The classifying work of products and services already exists. In fact, there are standardized classification systems in the U.S. and the EU. That is not the problem.

shop owners could opt-in or add additional rules to the algorithm.

No, they can't. Because there is no one algorithm to calculating the tax applicable to a transaction. There are thousands of different algorithms, many of which get changed on a yearly basis (and in some cases, even within a year due to special events like disasters). Each tax jurisdiction has its own rates and components. Within a single county, you can have different cities levying their own separate sets of factors and rates when determining sales tax. (See, for example, LA vs Santa Monica for examples of different rates on the same components, or any city in Colorado vs any other city in Colorado for examples of cities within the same county using different components and different rates.)

I think the 20% could be squeezed and solved a number of different ways especially in the name of the public.

Quite frankly, the storefront component is the 20%. You would need to devote 80% of your time to the tax component.

Note: in the "name of the public" the EU has introduced the VATMOSS system to simplify filing VAT reports throughout the EU, and the Streamlined Sales Tax program in the U.S. aims to do the same for filing US sales tax returns. However, filing sales tax/VAT returns was always the easy part of all of this; the hard part has always been figuring out the appropriate sales tax to apply to a transaction.

j / k navigate · click thread line to collapse