New Constructs was founded to build best valuation models in the business across thousands of securities. The old construct for models relies primarily on analysts who cover a small number of companies and independently build models that don’t all work the same way. To build the best models, we had to create our own data collection technology to ensure we had both complete and accurate data so that our clients could trust our research.
Back in 2003, our original customers included top portfolio managers at Fidelity, Janus and TIAA-CREF. These highly sophisticated investors could not risk making billion-dollar decisions with models built on traditional databases like Compustat or Capital IQ. They needed to know an accounting expert had analyzed every page of a company’s SEC filings to close GAAP accounting loopholes, and a finance expert had built a truly comprehensive valuation model.
Our machine learning technology has processed over 120,000 filings since 2003. Recent advances in natural language processing (NLP) enable us to leverage machines to automatically parse 85%+ of filings and footnotes compared to just 30% a few years ago.
CEO David Trainer explains the background and workings of our technology in the video below.
The keys to the successes of our data collection technology are:
- Experts in accounting and finance drive the machine learning and data parsing processes
- Vast knowledge base (120,000+ filings) of expert-verified parsing instructions informs the natural language processing technology
- Expert-verified models in the hands of sophisticated investors have vetted the results of our parsing and models for nearly a decade and a half
Value Investing 2.0
We provide the due diligence advisors and portfolio managers need to fulfill fiduciary duties on 10,000 securities. We make true value investing practical again.
Figure 1: Value Investing 2.0: Our Technology Offering In Pictures
Sources: New Constructs, LLC and company filings
For more insight into the technology behind New Constructs, we’ve created a page that details how we use source data from SEC filings and how that data is then analyzed and integrated into our company models.
This article originally published here on December 1, 2016.
Disclosure: David Trainer, Kyle Guske II, and Kyle Martone receive no compensation to write about any specific stock, sector, style, or theme.