Alpha Vertex Launches Alta to Create Investment Signals From Our Speech, Chaos and Language

By Mutisya Ndunda, CEO of Alpha Vertex

Part I of our series on Alta and using alternative data to gain an advantage in the markets.

While quantitative investing is all about numbers, the fact is that most of the world’s information is in written and spoken language and not encoded in sterile numbers.

At Alpha Vertex, we see this as a huge untapped opportunity and we have built advanced natural language processing (NLP) and machine learning tools to extract unique, high-value information — such as investment signals — from unstructured, text-based datasets with broad coverage.

Our team has been working hard on Alta, a new alternative data service, and we’re happy to be launching it this week.

It started with a simple, no brainer idea. There are 300 earnings calls that happen any given day during earnings season, but as important as they are — and as valuable as they could be to uncovering untapped insights into the company’s further direction or future failure — the truth is that nobody can attend all of these earnings calls.

Alta quantifies the information provided by companies in earnings calls, investor presentations and product announcements. The dataset includes over 140,000 conference call proceedings for more than 7,000 companies, starting from 2008 to the present.

A unique feature of our dataset is that the transcripts are in JSON format, which makes it possible to separate out different parts of the call. For example, it is possible to extract the prepared remarks by the company’s CEO or a CFO’s responses to analyst questions.

Built on our advanced NLP and machine learning solutions, Alta improves significantly on current methods of analyzing and delivering unstructured datasets. We provide all the raw data in a structured JSON format along with numerical features and indicators which summarize and quantify what was said and how it was said. Some of the topics that are systematically monitored and measured include:

• What executives say about the company’s earnings, financial outlook, buybacks and dividends
• Discussion of the macro economy, customers and regulatory developments
• Mergers and acquisitions

Additionally, we track all individuals who participated during the call and map them to a unique person ID which stays consistent even as individuals change roles or move between different firms.

For example, Tim Cook the former COO and current CEO of Apple, has a unique ID which is used to track his NLP features over time and across his two C-level roles at Apple. Along the same lines, we can track the Q&A asked by a particular analyst, for example Guy Moszkowski, from his time at Sandford Bernstein, Merrill Lynch and Autonomous Research.

Alta can save portfolio managers and research analysts considerable amounts of time by analyzing language complexity, summarizing tone/sentiment, diagnosing management personalities and extracting key topics.

The raw data and extracted features have broad applications including alpha signals in systematic strategies, early warning indicators for risk managers and bond investors, behavioral indicators for fundamental investors and competitive intelligence tools for corporations.

In fact, according to research from Columbia University, half of buy-side consumers listening to, or reading the transcript of an earnings call do not hold the firm’s securities at the time that they’re doing so. “We find that much of institutional buy-side consumption [of earnings calls] actually arises from non-holders who may consider [making] an investment or who use the call to fulfill other informational needs,” Heinrichs, Park and Soltes write.

OVERVIEW OF THE DATA

Earnings calls give company executives an opportunity to discuss their financial performance and answer questions from analysts and investors. The conference calls generally begin with prepared remarks followed by a Q&A. Prepared remarks are likely to be well rehearsed, or even prerecorded, but dialogue during the Q&A portion tends to be more natural and unrehearsed.

Remarks made by company executives during these calls serve two primary purposes:

• Informational: Provide information about the financial condition of the company
• Promotional: Manage external perception of the company amongst investors and analysts. For example, Zhou (2014) shows that executives play a blame game during conference calls by attributing success to internal factors and attributing poor performance to external factors, such as weather and economic environments

Our research suggests that features from both portions of the call are valuable and provide fruitful ground for alpha generation.

This post is a three part series, in a couple of days I’ll delve a bit deeper into the technology and tools used to build Alta.