4 min read

I Used AI to Summarize Nike’s Quarterly Investor Report

Anyone that takes investing seriously knows there’s a goldmine of valuable information packed into Quarterly and Annual Earnings Reports. But you also know how long and dense investor reports can be. Furthermore, when multiple reports drop on the same day, it’s difficult to find the time to stay ahead of all this new information.

That’s why natural language generation (NLG) models have such a vast opportunity in the investment world. One use case is using AI to summarize Quarterly and Annual Reports. With AI, we could quickly and accurately summarize investor reports in an easily digestible format for investors. This would save time and resources for investors looking for a comprehensive overview of a company's performance.

Kris Kay showed how he used ChatGPT to analyze and summarize Netflix’s most recent quarterly earnings report in the video below:

For a simple synopsis of a company’s health, I’d say that Kris’s experiment worked. It’s actually a great example of how to educate younger audiences on the markets and investing. But most of us don’t necessarily need language to be simplified. We just need the most valuable information extracted from the report.

My AI Investor Experiments

Equally as dense (and usually boring) is the quarterly investor call. These things can run for hours and contain more buttoned-up PR language than anything else. It takes true skill to be able to parse through this information for what’s truly impactful.

That’s why I used AI to extract the most valuable information from Nike’s FY 2023 Q1 Earnings Release Conference Call. Because there’s a lot of variability in how AI writing programs operate, I tested two different platforms/methods.

Test 1 – Generate A Summary

The first experiment was with LEX, a word processor with a generative AI writer feature. I copied and pasted the call transcript into the document and asked it to summarize the ten most important points. Here’s the result:

  1. Nike’s Q1 earnings beat Wall Street expectations, generating $10.2 billion in revenue and $0.57 in earnings per share.
  2. Nike’s digital sales increased by 82% and accounted for a third of the company’s total sales.
  3. Gross margin decreased by 90 basis points, primarily due to inventory liquidation in North America.
  4. Nike saw strong growth in China, with double digit sales growth and net income growth of 10%.
  5. Nike entered into a strategic partnership with Microsoft, establishing innovation “Fuel labs” that will further enable digital capabilities.
  6. Digital membership grew to 135 million, with members representing more than 74% of digital sales.
  7. Nike launched the Nike Plus App and is expanding its global Direct-to-Consumer offering.
  8. Late product arrival due to North American disruption is causing seasonally out of relevance inventory to be more aggressively liquidated.
  9. Nike expects a 150 basis point annual impact due to inventory liquidation, with the largest impact in Q2.
  10. Nike is confident that its five-year macro outlook is achievable.

What I love is that LEX did a great job extracting the important data while also keeping a few of the qualitative statements in there. What I don’t love is that there’s no feature to upload an entire document to be analyzed and I’m also skeptical of what it generated.

Test 2 – Extract Info From A Document

The second test I ran was with a NLP (natural language processing) platform called Upword, which promotes itself as a tool for summarizing long written texts. The result is that it pared down a 30-page transcript into a 4-page summary, greatly decreasing the density of the report.

You can read the summary here.

What I love about Upword’s result is that it kept all of the language intact. Whereas LEX generated new sentences with important information, Upword just quoted entire passages it found important. What I don’t love is that, in comparison to LEX, I can tell Upword passed over a lot of the data I’d like to know.

Endless Opportunities

Summarizing investor reports is a clear efficiency boost. But there’s a lot more potential for applying AI to the investment landscape in nuanced ways to extract information, not just condense it down.

For example, Gary Licht published an experiment using NLP (natural language processing) algorithms to find companies in weak positions that he should short.

Basically, Gary used NLP to analyze the language used in the last 20 years of public financial reports to identify what words or phrases can be predictive of a company’s collapse. His theory is that negative or worrisome words will be common across companies’ reports before they go into bankruptcy or experience a serious hit to their stock price.

Although Gary is still fine-tuning this investment analysis model, it shows the variety of ideas that will take shape with AI-assisted investing.

The two greatest challenges of this AI vertical, in my opinion, are:

  • Accuracy – Is this AI writing truth or fiction? With LEX, I found myself still searching the transcript to make sure LEX’s summary used actual data from the report.
  • Efficacy – Is this AI finding the best info? With Upword, I found it often extracting unuseful info that I normally would have just glossed over.

Having worked in venture capital and with institutional investors, I’ve seen the difference it makes when an investor knows the financial health of a company by paying close attention to dense reports. And I see how AI could help level up everyone’s abilities here. But I think this is an area of AI-generated writing that still needs much improvement.