This one’s for all the due diligence fiends and competitive landscape mapping mavens out there.
PitchBook, the data and analytics service for private equity and public markets, is rolling out an automated suggestions feature for premium users when they’re doing searches on companies for market intelligence.
The new service is based on machine learning technology that scours PitchBook’s financially focused information and data set. Each word in a description is represented in 300 dimensional space using the global vectors for word representation software lifted from researchers at Google and Stanford, and those vectors are then applied to companies to determine their various relationships.
“The differentiator for why the output of this is going to be high quality. When we look up a company is because we have this proprietary set of financial related news and information,” says Tyler Martinez, the director of software engineering and data science at PitchBook.
During an advanced search, the Suggestions algorithm stores the entire search as a vector ad compares it against a larger word embedding model to find similarities among companies.
Behind the new features is a years-long effort to get more financial data at more scale, according to the company. PitchBook invested in web mining tools and an automated news collection technology that can process 30 billion words.
And the amount of material that PitchBook and its competitors have to track has expanded exponentially since the company was initially launched years ago. There was $28 billion invested into 1,700 deals across the globe in the first quarter of 2018, and the geographic expansion of the private equity business and the explosion of interest in private markets has created a new demand among investors who don’t know what they don’t know, according to PitchBook.
“We built suggestions because it’s really hard to keep tabs on what is a big challenge in the market,” says Jenna Bono, a product manager for the company.