blog Market Intelligence /marketintelligence/en/news-insights/blog/unlocking-the-full-potential-of-earnings-transcripts content esgSubNav
In This List

Unlocking the Full Potential of Earnings Transcripts with Kensho NERD


Banking Essentials Newsletter 2021: December Edition


Automating Credit Risk Surveillance Using Statistical Models

Case Study

Searching for Alpha with Textual Data


Post-webinar Q&A: Speed and Scalability – Automation in Credit Risk Modeling

Unlocking the Full Potential of Earnings Transcripts with Kensho NERD

Most of the 3,500+ U.S.-based, publicly traded companies listed on major exchanges hold conference calls shortly after announcing quarterly earnings. The transcripts can be a rich source of information, often revealing insights not readily apparent in a company’s financial reports. However, the volume of information can be overwhelming for any analyst that needs to follow a portfolio of companies. There are several challenges that make it difficult to zero in on relevant information. This blog discusses three of these and introduces a solution for quickly identifying transcripts worthy of additional attention. 

Reviewing Corporate Transcripts Can Be an Onerous Task

Reviewing transcripts can be a time-consuming process given the following three challenges:

  1. Companies may be mentioned in numerous transcripts, not just their own. As shown in the example below, in the first quarter of 2020, The Boeing Company was mentioned in discussions by airline companies, suppliers, and other less obvious firms.

    • American Airlines: “The financial cost of the MAX grounding should be borne by Boeing shareholders, not American. We're pleased that recently, we reached a confidential settlement with Boeing to compensate American for the financial damages we incurred in 2019 due to the grounding of the MAX…”
    • Reliance Steel: “Beginning early in the fourth quarter, we proactively reduced our related inventory and headcount well in advance of the production pause announced by Boeing in December.”
    • Albany International (an industrial machinery firm): “While there are obviously near-term challenges driven by the ongoing grounding of the Boeing 737 MAX fleet, our longer-term vision and objectives have not changed.”
    • Columbia Banking System (a regional bank): “…I was wondering if you could provide some local color on the pause of the Boeing 737 MAX assembly at the Renton factory nearby. What kind of impacts you might see to the local economy…?”
  2. Companies can have names that include commonly used words. It isn’t enough to search a transcript text for specific words, it’s also important to determine if any mention refers to a real entity. For example, searching for “Target” in earnings calls from April 1-June 30, 2021 yielded 133 mentions, but only 30 of these were actually speaking about the chain store.
  3. Companies are often referred to by acronyms, ticker symbols, or previously used names (when doing historical analysis). For example, “Standard and Poor’s” may be called “S&P” or may be mentioned under the older parent firm names, “McGraw-Hill” or “McGraw-Hill Financial”. 

Now You Can Make Transcript Reviews Easy and More Insightful

Kensho Named Entity Recognition and Disambiguation on Transcripts (“NERD”) is a cutting-edge machine-learning system that unlocks the full potential of textual data found in S&P Global Market Intelligence’s Machine-Readable Transcripts. NERD identifies all companies that were discussed in any way in a transcript and where the mentions occurred. It then links the mentions to the appropriate S&P Capital IQ company ID, augmenting the textual data by making connections with other sources of structured knowledge for deeper insights, such as data provided through XpressfeedTM and Snowflake.

NERD is a probabilistic framework. Each possible match is given a score to indicate the likelihood of the text being an entity (i.e., the “NER score”) and the confidence that the link to the company ID is correct (i.e., the “NED score”). This enables users to identify cases with the highest quality links. It’s also possible to expand upon this and identify any call where a company may have been mentioned. For example, in the second quarter of 2021, there were 63 high-probability mentions of IBM, and 13 additional lower-probability mentions.

NERD is not used for alpha generation like Market Intelligence’s Textual Data Analytics, which generates sentiment scores and behavioral metrics based on company transcripts. Rather, NERD is used to quickly identify transcripts that are worth further investigation − whether for a fundamental analyst, investor relations department, or machine-driven applications.

Using NERD to compare the number of mentions for major automakers during the first half of 2021 versus the first half of 2020 revealed that these firms were discussed 38% more frequently in 2021. Although this doesn’t suggest anything positive or negative, it does say there was something going on that required further consideration. With the earlier Boeing example, the company was mentioned 133 times in the first quarter of 2020, but there were only 86 mentions in the first quarter of 2021 when the grounding of the 737 MAX was less newsworthy.

By using NERD in conjunction with Market Intelligence’s Machine-Readable Transcripts via Xpressfeed and Snowflake, users can quickly assemble a comprehensive list of mentions and then broaden the picture by linking to an extensive set of data. This can help reveal and describe less obvious connections between different firms, yielding a much richer understanding than if you only focus on the subject-company’s own calls.

Learn more about NERD on Machine Readable Transcripts
Learn more