Alex Rapp, co-founder of Calcbench, is working with his partner, Pranav Ghai, to put the finishing touches on the company’s open-source application intended to enable consumers of public company financial information to use and analyze data provided through XBRL. Calcbench won the inaugural XBRL Challenge award earlier this year based on the strength of its user interface and analytics, which allow users to perform multi-company comparisons annually and quarterly.
Can you give us an update on Calcbench?
We built this platform, and then we quickly realized it had some shortcomings in structuring data across multiple companies and industries. So we stopped working on the front end and have put in a tremendous amount of time working on the back end -- the way the database works, how it normalizes data points, how it catches errors and fixes them automatically. It has some pretty cool artificial intelligence things going on. There are a number of process that help catch and fix errors.
Recently we’ve turned our focus back to how people can use all of this cool stuff that we’re working on. We’ve made some improvements to the speed of the website, and we are building benchmarking tools. This is a new way to access information, compare information across industries and companies, and do valuations. It will also allow users to spot trends and to do some benchmarking and economic analysis. It will be fairly basic at first, but it will build and grow over time. We are hoping to launch it in the next few months.
How are consumers of XBRL data using that information now?
Well, that’s the thing -- there aren’t that many choices. You can bring the instance documents on to your desktop yourself, but that’s a waste of time really. You might as well just use the paper documents. We want to provide two different entry points -- a company-specific entry point where you can look at a company in detail, and then an industry or sector. You want to benchmark against other companies or a peer group. That’s the stuff we’re working on now. That’s the work that a lot of people are currently doing by hand.
Investors aren’t really going to use this for investment decisions until it contains three-plus years of history. One year of data in XBRL doesn’t do them much good, but it’s very useful for other purposes -- for people who are doing trend analysis or benchmarking. There’s a lot of information in there that’s very valuable, but we can’t expect every individual to understand how to use it just yet.
You’ve said you spend a lot of time correcting errors. What kind of errors do you see most commonly?
We’re fixing thousands of errors, some automatically and some by hand. We see cases where companies have their entire universe of numbers off by 1,000. It’s a scale error. They’re accustomed to truncating for presentations, but when you create an XBRL filing you don’t do that. A lot of the errors we see are not significant, but some can be very complicated. Sometimes we don’t even know for sure what the intended scale was. Take share counts, for example. It’s ridiculous how many companies get their share count wrong. It’s a little unsettling.
It’s also a schematic problem. There are still a lot of companies following a two-step process. They go through their usual close and filing, and then they hand the paper to someone, often a contractor, to be filed in XBRL. Hopefully that changes over time and people start making the XBRL filing part of the filing process. I’m pretty sure that extra step is the cause of a lot of these mistakes. Preparers just need to double check their work. The Securities and Exchange Commission will eventually lose its patience with these kinds of things.