Rating Systems Explained By: David Trainer

12 Mar 2026 · 1 min

Watch on YouTube

1:08

David Trainer explains how corporate filings have exploded from 25 pages in the 1990s to 200 pages today, mostly footnotes and legal language. His firm built a machine-learning system since 2003 to extract hidden gains and adjust reported earnings, but verifying these adjustments remains nearly impossible to audit.

Chapters

This is, this is like your average filing, right? And we know that financial statements maybe take up 5 pages. You've got risk factors and MD&A, another 20. So 25 pages for stuff. And by the way, when I started in the business in the mid-'90s, most filings are around 25 pages long. Now they average 200. Most of that's footnotes, 175 pages of footnotes, legalese. That's where we go to pull out unusual gains and charges to adjust the earnings number. And New Constructs is very simply a truth engine for earnings. We created machine learning capabilities going back to 2003. We started building this to go through these really long, complex, full of legalese and finance jargon filings to get through the footnotes and the fine print and pull out the information needed to get to the truth about earnings. And it's one of those things that you can talk about and everybody loves to talk about it. At the end of the day, you have to kind of put up or shut up as to whether or not your earnings adjustments work, because there are a lot of people that talk about earnings quality and earnings adjustments. And it's hard to know whether or not it's real. It's hard to audit whether or not somebody's reading financial statements, just like it's hard to audit your English class. The teacher knows whether or not you read the book.