This Software Can Uncover the Questions CEOs Don’t Want to Answer
Amenity Analytics’ founders say their platform can even pick up on potential deceptions on earnings calls.
(Bloomberg Businessweek) -- For as long as companies have hosted quarterly conference calls, analysts and journalists have tried to parse not just what a company’s executives say, but how they say it. Software from Amenity Analytics promises to automate this process by spotting when chief executive officers try to duck tough questions. The software, its makers say, can even pick up on the signs of potential deception that CIA and FBI interrogators look for—including stalling and the use of qualifiers—and can gauge the sentiment of what is said on calls and reported in public filings, issuing a positive or negative numeric score. The goal is to make it easier for investors to wade through information and quickly make trading decisions.
Amenity has just secured an additional $18 million in venture capital funding from investors that include former AIG CEO Maurice “Hank” Greenberg as well as insurance company Allstate Corp. Existing backers include Intel Capital, the venture capital arm of the chipmaker, and Israeli venture capital firm State of Mind Ventures. Amenity had previously raised $7.6 million in 2017.
The Benefit
Barclays Capital, in an October 2018 research report, concluded that using Amenity Analytics’ software to examine earnings calls—and using the sentiment score to decide whether to buy or sell short the shares of those companies—produced an annual return 12.97 percent greater than a benchmark index. The Barclays researchers also found that Amenity’s sentiment scoring, which is based on an understanding of words in context, was superior to a more common, cruder way of gauging sentiment that’s based on looking at the words mentioned most frequently on a call. Stock research firm Evercore said in a November report that it successfully used Amenity’s software to examine S&P 500 earnings calls for such macro trends as whether cyclical industries were more worried about a U.S.-China trade war and whether CEOs in aggregate thought the long bull market might be coming to an end.
Innovator: Nathaniel Storch, CEO and co-founder of Amenity Analytics in New York
Storch, a former Wall Street analyst and portfolio manager, knew from experience the difficulty of sorting through the information available on thousands of publicly listed companies to try to find trading signals. He teamed up with Ronen Feldman, a professor at Hebrew University in Israel who’s a leading expert in using software to extract meaning from text; together they formed Amenity Analytics in 2015. “We find all sorts of nuances in language,” Storch says. That includes certain tells that indicate a person may be lying. “Stall tactics, redirects, evasive answers, the use of clichés and jargon—there is a whole taxonomy of factors that you can see,” he says. Amenity’s software can be used to further research any industry, but Storch and Feldman decided to start with stock research and insurance since they understood the need most clearly.
Challenges
Storch says the software works best when a user compares a call’s overall sentiment score or the number of instances of apparent deception to the figures for prior calls, or benchmarks the data across an industry. After all, some CEOs are just more prone to using clichés—which could be a sign of evasion or might just be the way that CEO talks.
Amenity has lots of competition, from International Business Machines Corp. and its Watson AI platform to Alphabet Inc.’s Google, as well as from similar analysis systems built by some large banks.
The Verdict
Amenity’s software is used by ratings agency Moody’s, the Nasdaq, and media giant TimeWarner, among others. When I tested the software on the transcript of the infamous Tesla quarterly earnings call from May 2018 in which Elon Musk berated an analyst for asking “boring” questions, the software scored the call as off-the-charts for “deception.”
Then again, analysts probably didn’t need a piece of software to tell them that.
To contact the editor responsible for this story: Dimitra Kessenides at dkessenides1@bloomberg.net
©2019 Bloomberg L.P.