Reimagining software for equity research

Summary

This post explores building an investment platform  that helps retail investors analyze a business for investment purposes. While this approach can be extended to other asset classes, the main focus here is on businesses as investments. This memo will go through the current state of equity research platforms and what problems this new proposed platform aims to solve.

Customer

For the core thesis, we will primarily focus on retail investors with a long time horizon. They generally look for high quality businesses with good fundamentals. They also attempt to limit downside risks by developing a deep understanding of the business. This persona generally believes in the effects of long term compounding by holding on to investments for multiple decades. Value investors adhering to the Buffett philosophy of finding ‘wonderful’ businesses at a fair price typically would fall in this category. Dividend investors looking for consistent passive income for extended periods would also be a representative persona.

While short term traders and institutional investors can also benefit from this platform, catering to their needs may require slightly different solutions that might be beyond the initial scope of this platform. Similarly, long term passive investors who buy an index would not fall into this category.

As these investors get more sophisticated and develop their circle of competence, they rely on a set of different resources to identify new investments and monitor them. Typically buy and sell decisions are rare, but require deep due diligence. Here is a common set of resources they use for these purposes:

1. Social media sites such as Twitter and Youtube
2. Annual and quarterly reports from companies
3. Apps such as Seeking Alpha and Simply Wall St for stock screening purposes
4. News outlets such as Barron’s and Wall Street Journal
5. Brokerage account for price and performance data
6. US Department of Labor or St. Louis Fed for economic data

Primer on research

While institutional equity research can be fairly complex, we will try to break down the research process from the perspective of our target customer. Such research usually involves both quantitative and qualitative analysis. We avoid using the term fundamental analysis which traditionally leans heavier on quantitative analysis, specifically around valuation of a business. The evaluation of qualitative aspects of the business may not be traditionally emphasized in fundamental analysis. 

A quantitative approach relies on analyzing the company’s financial statements to make estimates of future earnings. A qualitative approach relies on understanding of the product or services that the business sells. Some relevant factors that impact such analysis are the competitive landscape of the industry, macroeconomic headwinds/tailwinds around the business, and management’s strategy including decisions around capital allocation. While these areas are not cleanly segmented, they help an investor make estimates and judgments around the companies ability to generate returns for the investor.

Typically, such investors are attempting to discover and monitor high quality businesses. While there is no standard definition of a high quality business, here is a concrete example of how Bill Ackman, an institutional investor and long term Buffett follower, would define attributes of such a business:

• Simple, predictable, and free-cash-flow-generative
• Formidable barriers to entry
• Limited exposure to uncontrollable extrinsic factors
• Strong balance sheet
• Minimal capital markets dependency
• Large capitalization
• Attractive valuation
• Exceptional management team and governance

The assessment of attributes such as strong balance sheet or attractive valuation can be performed using quantitative analysis of financial statements. However, attributes such as formidable barriers to entry, limited exposure to uncontrollable extrinsic factors require a deeper understanding of the business model and the sector/industry it operates in. Even evaluation of the management team and governance needs requires review of past decisions such as buybacks or share issuance and track record of disciplined execution over an extended period.

Problem

The average retail investor is getting more sophisticated as financial data about businesses has become more accessible. But, they still come across a few problems:

Lack of objectivity: Financial information is almost always layered with opinion to attract more consumers. This leads to exaggerated interpretation and reporting of the facts. Generally, incentivized by clicks, news sources tend to provide articles on a spectrum of dire circumstances to unbridled optimism. Trusted sources like SEC filings by companies help paint a more accurate picture. But, the verbosity of such filings make it hard to extract relevant information. While most apps enable you to perform quantitative analysis based on financial statements, people rely on news, social media, or hobbyist analysts to gain an understanding of the qualitative aspects of the business. We can certainly do better.

Financial analysis requires expertise: Retail investors are at different stages in their abilities to analyze a business. There is a lot of good  quantitative and qualitative information in financial reports and investor relations pages of companies, but it could be difficult to interpret. Also, an untrained eye can miss anomalies that are sector or industry-specific. The pattern recognition of anomalies or positive trends comes from years of evaluating a sector and deep understanding of the GAAP accounting metrics of such businesses. It’s also difficult to come up with a checklist of items, so it’s easy to miss relevant aspects that lead to poor decision-making. While data aggregation seems like a solved problem, the synthesis of that information and discovery of interconnections between datasets is left up to the investor.

Accounting for extrinsic factors: You can think of financial analysis as a set of signals that aid with the prediction of future prospects of a business. Factors such as interest rates, currency fluctuations, supply chain shortages, commodity price fluctuations can affect earnings of a business. However, these interconnections with extrinsic factors are rarely modeled by any existing solution.  In 2022,  retail companies failed to accurately model demand from consumers leading to overstocking. Conversely, the supply chain shortages in 2020 resulted in higher shipping costs/times for companies leading to delays in product deliveries. The ability to understand these economic interconnections and their impact to businesses can help investors make prudent buy/sell decisions.

Mission

Our mission is to make it easy to analyze businesses

We believe that to truly understand a business and its earning potential, you need to be able to synthesize the impact of both intrinsic and extrinsic factors that affect a business. While making this information accessible to an investor would be valuable, we will always strive to go further and help them make sense of the data needed to analyze a business. We will help connect the dots between the published metrics about the business and any information that explains fluctuations in them. We will also help the investor understand how the business is doing with respect to its peers and what areas stand out as competitive advantages. Additionally, we will try to model any extrinsic economic factors such as supply chain disruptions, weakening demand, interest rates, or any possible correlated factors that affect a business performance. We will also strive to provide investors with accurate and updated information as close to the source as possible.  

Vision

Our vision is to model the global economy and all the participants that transact within it.

A business is a participant that provides goods or services within a context of a global economy. If we were to visualize the economy as a graph, then all the businesses and customers would be nodes with transactions as edges.

The diagram below is intended to express such connections. While this is far from comprehensive, it is meant to help visualize this approach:

If we are able to generate such a live graph, we would have mapped all such interconnections within a global economy and visualize the impact of any of these transactions in real-time. Assigning weights to these edges (or a set of edges) can also help us understand the impact of a connected nodes on a business. We can start small by modeling nodes such as businesses. But, as we expand our understanding of all the extrinsic factors that impact transactions, we start uncovering correlations within the economy. For example, the central banks have the power to influence rates and increase or decrease the number of transactions within an economy. Lower rates result in a decrease in consumer spending, which has a direct impact on revenues for businesses. Conversely, higher rates result in lower demand, whose impact can be seen on retail sales, advertising spend, and even freight rates. So, the global central banks should also be modeled as nodes in this graph.

If we take this further, we start realizing that asset classes have correlations too. Rising rates impact housing demand. Home-builders curb supply during these periods which results in lower revenues for such businesses. But, sellers also have a tough time finding buyers resulting in lower housing prices. Even commodity price fluctuations impact the bottom line of certain businesses. For example, the price of lumber affects the revenues of businesses in the home improvement sector. 

Success would be a queryable live graph of the global economy which helps anyone easily understand the implications on each participant as events happen. Investors would be able to make more informed decisions about their investments. Businesses would be able to better plan their inventories, manage their costs effectively, and even perform competitive analysis to build better products or provide better services. Central banks can make more informed decisions about federal rates. Governments can make better economic policy decisions. Academicians can leverage such a knowledge graph to perform deeper economic research.

Product

The proposal is to build a multi-platform app that is available to the investor at all times. The initial goal would be to search for new investments and monitor existing ones. The core focus would be to provide a research platform that goes beyond data aggregation and finds correlations between different aspects in the economy that affect a business. Even, within a business, we will use technology to really understand how the business is performing. While we plan to leverage standard GAAP and relevant non-GAAP measures to show the quantitative details, our focus will be on qualitative insights about the business, the sector it operates in, the competitive landscape including durability of cash flows, and the extrinsic factors that impact its operations.

Modeling businesses from SEC filings

In terms of the nodes, let’s just start with businesses. A good place to start would be to make it  easy to extract all the content in filings and present it in a consumable manner. Here is an example of how income statements of a business  could be made more consumable by the use of Sankey charts.

A quick win would be to just attribute a significant increase/decrease in a GAAP metric to a reason:

Additionally, we could leverage document extraction techniques to extract key sections from filings.  Most apps just present a high level company overview, but fail to describe the products or services that the business provides. Another quick win would be to just extract this info and make it more accessible.

You could imagine an organized breakdown of revenue and costs based on these products within the app. Breakdown by segment would be relevant too.

Doc extraction  can also help with extracting valuable charts like this from the filings.

Similar to Blinkist for books, we could also consider summarizing these reports using AI or simply just extracting relevant aspects from the reports. Picking out salient aspects of an earnings call could also be good use of such summarization technology.

Modeling the Economy

The first phase would be to aggregate all the important economic datasets – CPI, employment rate, M2 supply, GDP, federal funds rate, etc. Then giving the user the ability to overlay this economic data on any GAAP or segment/product specific metric for the company. This could also help us identify leading and lagging indicators. It would help the investor better understand business cycles.

Here is an example of how commodity prices overlayed on revenue, paint us a picture about margins:

The real value comes up from being able to determine what extrinsic factors affect each business and by how much. Once we account for these variables, our estimates get better in predicting business cycles.

As we start recognizing patterns and understanding correlations, we can start sending insights based on anomalies. We can also let users set alerts based on anomalous trends. There would also be value in letting users easily query this data directly to screen for stocks or understand trends. Essentially, it would become a data warehouse for joining both economic datasets with financial metrics from businesses. 

Datasets such as credit card transactions and home sales can also be seen as incoming signals that can be leveraged to train ML models. These models can be leveraged to make predictions on earnings and demand cycles. 

Consumable Content

Community tends to be an important aspect when it comes to discovery of information about businesses. There is value in aggregating all social content about a business and ranking them based on trust scores. We should also provide tools to help users determine and verify any information within the content they are consuming. See commentary section below on battling misinformation.

There is also an opportunity to create custom informational content based on facts. We could also let companies publish such content for their investors. This would be a more of a centralized and standardized investor relations view of each business. We could provide a framework for creating short-form videos or templates for publishing information about products that PR/marketing orgs within these companies can leverage. 

Better Tooling

While valuation and screening tools could be helpful, they are easily available and we should just build them in for the sake of completeness. This probably shouldn’t be the core value prop of the platform. But, some areas to consider:

• Peer clustering for relative valuation – this should help investors both identify peers based on business attributes, but also value them relative to other companies within the sector.
• Discounted cash flows (DCF) for intrinsic valuation – DCF is a common valuation strategy that relies on leveraging estimates of future cash flows of a business. ML techniques can really help with coming up with better estimates.
• Advanced screeners based on Magic Formula/Acquirer’s multiple – There are advanced strategies that people have started employing for screening stocks. Providing backtesting functionality to test the performance of these strategies could be a valuable tool for investors.
• Data export – Another key feature could be to enable users to export financial data to spreadsheets or even provide a spreadsheet-like experience within the app. This is helpful for calculating ratios or estimating future cash flows.
• LLM-based data synthesis – Enabling large language models (LLMs) to perform qualitative analysis based on the entire corpus of filings and investor materials could significantly cut down analysis times. We need to be careful about not relying on LLM’s for deterministic behavior or mathematical calculations. But, it should definitely help with detecting blind spots for an investor.

Disclaimer on stock picking

Most retail investors are likely better off buying  S&P 500 or total market index funds. It is also important to note that retail investors are unlikely to have an ‘information edge’ over their institutional counterparts. But, retail investors have been stock-picking since the dawn of the stock market. Our goal here is not to provide investment advice, but help them become more informed about their investments and gain an understanding of macroeconomic conditions. See chart below of even institutional investors failing to beat the market with an information edge.  

Commentary

Battling misinformation in the media

Investors consume information through various sources such as Yahoo Finance, Bloomberg, and Social Media. However, all of these sources provide a layer of interpretation coupled with commentary from the author/creator. This makes it difficult to maintain objectivity and separate fact from opinion. The bias for sensationalism or hype is caused by the fact that social media or even traditional media is incentivized by reach and impressions. It occasionally puts the consumer of that information through a cycle of confirmation bias of a sub-consciously held perspective. However, when evaluating investments, rationality and objectivity must prevail. 

Even for businesses, misinformation can cause sudden swings in share prices. A sudden drop in share prices makes it harder to attract or retain talent. It also makes it challenging to issue shares to raise debt. 

Such misinformation in the media about Coinbase prompted Brian Armstrong to write this essay about Coinbase FactCheck. He starts with the statement:

Every tech company should go direct to their audience, and become a media company.

In principle, I agree with this statement. But, I believe every company faces this challenge, but it is certainly exacerbated in tech. Of the three options, Brian laid out, we aim to facilitate option 3 – publish the truth. This platform could eventually also become a way for businesses to reach their investors directly and build a relationship.

To be clear, our core customers are investors. Aligning our interests with them helps us maintain objectivity and integrity of the platform in the long run. By providing such a marketing tool for businesses, we begin aligning our interests with businesses that could dilute our credibility. We should tread carefully.

The problem of misinformation gets worse when you are dealing with international companies

Here is an example of a twitter exchange about Jack Ma ceding control of Ant Group. In this case, media outlets like Wall Street Journal or Forbes actually did not publish inaccurate information. The challenge, however, is that the articles paint an incomplete picture and investors jump to an invalid conclusion about Alibaba, a company that owns a stake in Ant Group. Here is a twitter exchange about this interpretation: 

In this case, both the original post and my comment were inaccurate. The author feeds into their confirmation bias that everything is going wrong with this company. My response feeding into my pain avoidance in knowing that there is likely a material impact to the future prospects of the business although the stake is intact. So, both of us are adding a layer of opinion to this. 

However, the quoted tweet is the actual piece of information that should be provided to the investor to derive an inference. In this case, the main piece of evidence that is missing is the fact the BABA’s stake in the business is not impacted by Jack Ma ceding control since his stake is through other entities. In order to find this information, I had to dig through the IPO filing prospectus for Ant Financial. This is where technology would have helped:

• Finding a “trusted” source like a document filed to a regulatory body.
• Indexing ownership structure information and surfacing that when researching this new development.