CIF Time Series Complex Event Analysis on Earning Calls speaks alpha to Investors and offers strategic edges to C-Suites

“Time Series Complex Event Analysis” analyzes a sequence of time based document to detect headwind and tailwind.

SiteFocus
SiteFocus
Sep 6, 2018 · 6 min read
Time Series Complex Event Analysis on what executives say in earnings calls can reveal headwinds and tailwinds impacting company performance

Summary

Applying “Time Series Complex Event Analysis” on Earnings Call Reports of a business reveals headwind and tailwind of a business operation. Such information is missing from a balance sheet or causal reading of Earning Call Reports. Qualitative analysis with Symbolic Logic on time series earnings call gives investors the big picture that is currently unavailable among all investment intelligence vehicles. By applying Symbolic Artificial Intelligence algorithms on a time series (a sequence of Earnings Call Reports from a business), investors can get the clarity of headwind and tailwind ahead of the business operation. This new analysis will give investors the vital information that is needed for alpha investment, equally, it will provide valuable insights for C-Suite officers to gain the competitive edge.

Introduction

Earning conference calls offers health and performance data about a company. Hidden inside earning call transcripts are a reliable source of information that provides a rich data platform that one can use to analyze a long list of driving forces relating to financial prospect of business sectors, micro and macro economic such as:

  • Business activities on a business sector
  • Outcome of a business operation
  • Trajectory of a business strategy over a period of time
  • Result on the pivots of business models
  • Supply chain conditions
  • Business models
  • Operational risks
  • Competitiveness within a business sector
  • Effect of regulations
  • Trading dynamics such as effect on tariffs
  • Labor and human resources
  • Capital expenditures

Qualitative Analysis of Earning Call Reports is still a job for human experts, financial institutions want to automate this process to no avail.

Quantitative analysis can tell if a company beats the expectation of Wall Street analysts by numbers, but it falls short in deductive rationale. Unfortunately, data from transcripts are in natural language and is beyond the scope of quantitative analysis. Acquiring the qualitative aspect requires qualitative analysis. It is the qualitative part that is hard to achieve. The voluminous number of transcripts from earning conference calls, over a period of time, make it humanly impossible for anyone to get the big picture.

Prevalent Solutions are inadequate

Modern text analytics are among the favorite tools used by investment firms to analyze unstructured textual data such as earnings call transcripts. Common approaches were to assess the qualitative aspect by means of vectors, word count, sentiment analysis, hashtag, word cloud, or semantic mapping. None of these analytics shed light on deductive rationale, causation, or time series of complex events. Analytics based on machine learning that offer insights based of past events cannot be used. This is because historical data cannot be used to model unknowns. In recent days, a growing number of expert opinions in the Artificial Intelligence and machine learning fields have converged towards the conclusion that data science based on machine learning or deep learning has limited use in natural language understanding.

A Solution that delivers Alpha Investment

Symbolic logic is a natural fit for qualitative analysis of complex documents written in natural language. The foundation of symbolic logic was formed over hundreds of years. It is a method of solving problems with deductive reasoning as opposed to most modern data sciences that rely on inductive reasoning based on probability and repeatable patterns. Our implementation of symbolic logic does not require repeat occurrences of similar data to make sense. In a nutshell, we have devised an algorithm that transforms complex scenarios into atomic formulas and connectives that are clustered inside semantic neighborhoods. This innovation enables us to break down complex scenarios in favor of symbolic logic, in the process, enabled us to automate the generation of rule sets in the form of symbol chains that can be classified into two groups labeled as headwind and tailwind. Analyzing the distribution of semantic neighbors among these two groups enables the user to understand and depict the causation of headwind and tailwind of complex scenarios embedded inside an earnings call transcript. This system is available for general use through SaaS subscription.


An example of using headwind-tailwind analysis on an earnings call

LULU’s extraordinary performance in stock price after its recent quarterly earning call in August raised curiosity on the causation of such behavior. For the purpose of demonstrating the use of CIF to analyze earnings call transcripts, we used the last three quarters of earning conference calls on Lululemon Athletica Inc. (LULU) as input data.

A snapshot of LULU’s 5-day stock chart Aug. 28, 2018 — Sep. 9, 2018. Source: MarketWatch.

We asked the system to identify the headwind and tailwind of the on-going operation of Lululemon with the following question:

“How does revenue, product, margin affect future outlook?”

CIF automates the process in 5-steps:

  1. Define a project and topic for CIF to store the transcripts as input data for analysis
  2. Ingest earning call transcripts into topics
  3. Run a time series analysis on the 3 conference call transcripts
  4. Compose a free form question to ask the system on subjects of interest: “How does revenue, product, margin affect future outlook and business model?”
  5. Run “Time Series Complex Event Analysis” to get the answer

CIF uses adaptive algorithms to discover unknown entities and relationships found in complex natural language documents. Unlike other textual analytics that requires dictionary, ontology, or pre-tagging, our system does not require any pre-processing. The basic setup comes with default settings. All that is required is to use a Web browser to step through these five steps to get the result.

The following is line graph of the headwind/tailwind analysis of the three quarters

CIF Complex Event Time Series on LULU Q4FY17, Q1FY18, Q2FY18 earnings calls. Source: SiteFocus.

Referring to the line graph, the blue line represents the tailwind and the red line represents the headwind. Over the past three quarters, we observe a gradual decrease of tailwind and a gradual increase of headwind. We also observe the operational trajectory of the business model undertaken by Lululemon over the course of 9 months.

Time Series Complex Event Analysis Report

The CIF system produces a detailed report which includes headwind and tailwind tables. The tables provide operational details in terms of symbol chains and associated excerpts from the analyzed documents. It is beyond the intent of this document to describe this report at full-length. A brief explanation follows.

As we can tell from the tailwind table, the symbol chain “men*product” is indicative of the main focus for the current quarter. The operational trajectory started two quarters ago from “business*digital” to “men*product”, showing a steady increase in optimism while there is increasing headwind on revenue. The following is an example of an excerpt from the detailed report produced by the system. The report provides supporting excerpts on all relevant symbol chains found in the Semantic Neighbor Distribution reports.

An example of a symbol chain and excerpt

A symbol chain (MARGIN -> MEN) and transcript excerpt automatically pulled by CIF. Source: SiteFocus.
An example of the headwind table.
An example of the tailwind table.

Closing thoughts

The big picture of an investment portfolio is comprised of qualitative and quantitative components. Quantitative analysis can only tell the story from the balance sheet’s point of view. Qualitative assessment reveals the fundamental strength through the depiction of business operational trajectory. This information flow enables investors to better calibrating business outlook with market dynamics. The advent of symbolic AI makes it possible to mitigate qualitative risk previously inaccessible to support investor decision-making. This is the first in a series that explores the use of Time Series Complex Event Analysis on earnings calls. Please feel free to reach out with questions or comments on this subject.

SiteFocus

Written by

SiteFocus

Pioneering #SymbolicAI solutions for natural language that help de-risk strategic decision-making. Also, #AI-on-the-#Edge. Visit: https://www.sitefocus.com

Welcome to a place where words matter. On Medium, smart voices and original ideas take center stage - with no ads in sight. Watch
Follow all the topics you care about, and we’ll deliver the best stories for you to your homepage and inbox. Explore
Get unlimited access to the best stories on Medium — and support writers while you’re at it. Just $5/month. Upgrade