Investment Office Logo

Devil’s in the Details: Robo Analysts, Financial Footnotes, and Buried Insights

"In all honesty, real fundamental analysis hasn’t mattered since before the tech bubble, but the industry is running out of ways to avoid doing this research. Consequently, what we’re seeing now is a return to fundamental research."
Prattle, December 14, 2017

Prattle is pleased to publish this interview with David Trainer, CEO of New Constructs, focusing on how machine learning technologies can add significant value to the fundamental research process and lead to better returns.

You have a reputation for producing detailed research. What has led you to emphasize detail in your approach?

To some extent, I’d say it comes naturally to me. But my experiences have certainly shaped my approach.

During my college years, I became disappointed with the lack of rigor I saw in academic finance. This disappointment carried over to my first job, where, as an auditor, I caught an error in a filing, reported it to my manager, and was informed that the issue would be fixed—in other words, buried—in a footnote. For my next position, I chose executive compensation consulting because the focus was on going to boards of directors and explaining, “whatever you do, don’t pay executives based on accounting earnings, because you can grow accounting earnings while running the business into the ground.”

As I continued to have experiences like these, it became clear that investment-relevant details about company performance were getting buried in dense notation or accounting conventions, and I saw this as an opportunity to create value.

When I began working for Credit Suisse, I was given the opportunity to put my observations into action. I built the most comprehensive, detailed model possible, based it on economic earnings instead of accounting earnings, and made it work for all companies across all sectors. Because I understood accounting idiosyncrasies and did thorough reviews of annual reports, my work was well received and, eventually, adopted by analysts.

While one can skirt around the details and still achieve modest success in this industry, the price you can pay for superficial analysis can be high. Think of the big scandals that have rocked finance—Enron, WorldCom, etc. These companies were able to rob shareholders and get paid handsomely to do so because they could hide their gross mismanagement of capital and nonsensical executive pay with arcane accounting practices.

Rigorous, detail-centric analysis may not be fun, but it’s vital to informed investment decisions. 

What are the hallmarks of high-quality investment research?

Fundamental research has a long history, and, in principle, analysts know how to do high-quality work. In many ways, the problem is not methodology so much as it is implementation. It’s just difficult to do everything high-quality research requires.

Over the last few decades, financial filings have exploded in length, from a dozen to hundreds of pages. In addition, the accounting language and calculations used to describe company performance has become increasingly opaque and, in some cases, downright unreliable. To make any determination about the health of a company, an analyst must diligently review all of this material. Eventually, cost/benefit analysis kicked in, and analysts abandoned deciphering annual reports for easier, more profitable work, like IPOs.

As time passed, it became commonplace to outsource fundamental research. While there’s no substitute for rigorous, traditional analysis, the influence nonfinancial factors have had on market pricing has allowed for this substitution. In all honesty, real fundamental analysis hasn’t mattered since before the tech bubble, but the industry is running out of ways to avoid doing this research. Consequently, what we’re seeing now is a return to fundamental research.

In your opinion piece in the Financial Revolutionist, you talked about the bright future “robo analysts” have in investment research. Could you explain your thoughts here?

To pick up where we just left off, what we’re seeing now is a return to fundamental research. Everyone in the investment space wants a true measure of a company’s profit, but most don’t have the time to actually ensure they get this information correct. Fortunately, technology is now at a place where largely automated systems, robo analysts, can do fundamental analysis with more precision while making it cheap, quick, and easy for financial professionals to have access to this research. And the industry is welcoming this innovation with open arms.

This comes at a particularly key time for asset managers, as the fiduciary rule has raised the bar for the quality of investment research they can use. Sell-side research, which accounts for the vast majority of investment research, carries too much risk of being conflicted and won’t hold up in this emerging regulatory environment. As a result, asset managers are in real need of alternative research solutions that provide an objective assessment of profitability and legitimate insights into the future profits required to justify stock prices—all at a reasonable cost. I believe robo analysis can do both of those things.

Could you talk a little more about your own process and approach—and the robo analyst space more generally?

The first step in our process is collecting the data. Basically, all SEC filings, every page, have to be thoroughly read and evaluated, and we use both human and machine processes to do this work. Once the data is collected and compiled, step two is feeding this data into models that convert the collected information into economic statements, and step three is quantifying market expectations based on our models.

One thing I’d like to emphasize here is the importance of quality underlying data to accurate results. One of the ways we’ve been able to distinguish ourselves in this space is the depth of our subject matter expertise in accounting and finance. That expertise drives our technology and allows us to scale it and transform unusable, unstructured information (like an annual report) into model-ready, structured data.

The level of rigor and sophistication we bring to this whole process has truly set us apart. As far as I know, our robo analyst technology is unrivaled. There are firms that do parts of what we do. We have competitors, for instance, that hire thousands of people in third world countries to manually comb through filings and collect the data, but such approaches inevitably produce flawed datasets. The data collectors rarely have strong financial training, English is often not their first language (try footnotes English), and it is not uncommon for clients to do the first real QA on the data. Don’t take my word for it…big 4 accounting firm Ernst & Young recently published a white paper proving the material superiority of our data versus Bloomberg and Capital IQ.

How will equity analysis change over the next 10 years?

One of the biggest challenges facing firms over the next 10 years is to cut costs while simultaneously improving research quality.

The monumental shift from active to passive investing in this industry isn’t going to stop, and it’s putting tremendous pressure on active managers to tighten their belts. At the same time, regulations like the fiduciary rule are forcing research quality to improve. This shift obviously puts firms in a difficult situation, and I think they should look to robo analysts to solve the puzzle.

About David Trainer
David is CEO of New Constructs, an independent research firm that leverages proprietary Robo-Analyst technology to find key insights from the financial footnotes of 10Ks and 10Qs. David is a distinguished investment strategist and corporate finance expert. He was a 5-year member of FASB’s Investors Advisory Committee and the author of the “Modern Tools for Valuation” chapter in The Valuation Handbook. David’s insights into the markets and his stock picks have been popular with a wide variety of media outlets.

About New Constructs
New Constructs leverages the latest in machine learning to analyze structured and unstructured financial data with unrivaled speed and accuracy. The firm’s forensic accounting experts work alongside engineers to develop proprietary NLP libraries and financial models based on the best fundamental data in the business for stocks, ETFs, and mutual funds. Their clients include many of the top hedge funds, mutual funds, and wealth management firms. New Constructs has partnerships with Thomson Reuters, Scottrade/TD Ameritrade, Interactive Brokers, and Ernst & Young. Ernst & Young recently published a white paper proving the material superiority of the firms’ s research and data.