The Dumbest Problem In Financial Technology

The deeper I get into this insane world of financial technology the more I find incredibly dumb things that need to be fixed, augmented, or completely disrupted. This is what happens when an oligopoly of companies (Thomson Reuters, Capital IQ, Factset, Bloomberg) basically dominates an industry for a few decades.

There is one specific problem that is so core to just about everything that it BLOWS MY MIND someone hasn't solved yet. It's a problem that we face at Estimize every day.

What is it?

Simply getting a reliable set of reported numbers from publicly traded companies each quarter which match up on accounting standards to previous quarters.

How important is this? Well, for just about every single trader, analyst, and PM it's crucial to have a clean set of numbers to compare against, to run screens against, to run algorithms against. This is about as crucial as it can get, and all of the large and small financial data companies seem to always fuck it up somehow.

I was out in Santa Barbara at the Goldman Sachs Innovators conference a few weeks back when Netflix reported their Q3 numbers. The stock got obliterated. Interestingly enough Reed Hastings their CEO was speaking at the conference the next day. So I decided to take a look at the report to see what went so wrong.

It would seem based on this chart of Netflix quarterly revenues and this quarter's guidance numbers that the guidance was why the stock got hit so hard. But upon closer inspection any rational human would question why Netflix's revenue growth would so abruptly screech to such a halt. So I did some digging.

It turns out that Netflix normally reports full revenue guidance, as in, all of their revenue. But for some odd reason, this quarter in the press release they decided that it would be wise to only cite "streaming revenue". Streaming revenue represents about 90% of Netflix's revenue, but every single financial data platform as well as a ton of quants and discretionary traders made the mistake of putting that number up as full guidance and comparing it to previous total revenue growth (about 27% YoY). And yea, if you saw revenue growth slip from there to about 12% YoY you'd sell the crap out of that stock too.

So everyone was using the wrong numbers and the stock got decimated. The next day some traders went back and looked at the actual release and saw the word "streaming" and quickly bought the stock. It went from $331 to $390 in less than two weeks, a nice little 20% gain for whoever was paying attention.

This shit happens ALL THE TIME.

This is a problem that can be solved!

I'm giving everyone a startup idea that is worth a good deal of money if executed well, and isn't that difficult to execute. Here's how it works.

1. Build the necessary web scrapers to pull all of the relevant financial info from the press releases.
2. Throw all of that data into an admin along with the actual press releases and conference call transcripts.
3. Build a Mechanical Turk job which asks people to go into the press release to find each important number and enter into the correct field.
4. Each release must be looked at by 3 people. If the numbers from all 3 people match, that's a clean number. If the numbers don't all match, it goes back through the same process again. If it fails the matching process again it goes into a que where a human analyst at your company reviews it.
5. You should now have a cleaner set of data than just about everyone on the planet with virtually zero errors.
6. Sell data via API and Excel plugin.
7. Give middle finger to other financial data companies.

The old way of doing this process is to hire 300 people in India who have no clue what they are doing to manually enter this data. Believe me, we see this at Estimize all the time, fat fingers, decimal points in the wrong place, or just people who don't take the time to verify the number is Non-GAAP vs GAAP.

Mechanical Turk would be a much more scalable, reliable, and cheap way to do this, and you wouldn't even have to employ anyone in India (though it's likely that most of the people doing this on Mechanical Turk are in fact in India).

Please, please, someone build this company.