In Supermoney, author Adam Smith asked Buffett what advice he would give his younger self if he were to start all over again to which he replied:
Buffett: […] if he were coming in and working with small sums of capital I’d tell him to do exactly what I did 40-odd years ago, which is to learn about every company in the United States that has publicly traded securities and that bank of knowledge will do him or her terrific good over time.
Smith: But there’s 27,000 public companies.
Buffett: Well, start with the A’s.
When I read this some years ago, it had a profound impact on how I approached equity research. Buffett’s advice might have to do with there being fewer options to scale equity research before the Internet so researching stocks meant flipping through print materials and physically requesting company reports, sometimes without luck. What Buffett mainly did when running his 50s and 60s partnership was flip through Moody’s manuals and ValueLine where each company’s key components spanned one page each.
This is a far cry from what’s possible in the Internet age. But while the Internet brought with it a lot of efficiencies it also brought a lot of laziness. To find a shining investment, Buffett had already become familiar with every investment possibility in the near universe through book flipping and report reading. In their attempt to find a shining investment today, modern investors use stock screeners, easily-accessed sell-side reports, charts, and an endless stream of other people’s opinions as to whether a stock is a buy or sell—all those short-cut activities that postpone rolling up the sleeves, doing the work yourself, and wrapping your arms around it. It’s so easy to learn something about pretty much any business by merely existing on the Internet, subscribing to some Substacks, and scrolling Twitter, but the vast majority of such information is secondary. Sometimes it’s wrong. The result is that most investment decisions come about from real-world opinions and nicely-told narratives that form ‘your own’ thesis without cross-examining your own thought process before pushing the buy button.
When Buffett bought Precision Castparts, he ostensibly did so by snap decision—however, a snap decision that was a result of accumulated knowledge not only about the company over an x period but a latticework of businesses like it and businesses unlike it through years and years. No alternative to hard work could have reached such a quick decision.
What we can infer is that there’s this paradox to excellent stock picking. Studying every company in the universe can give you the breadth of knowledge and pattern recognition you need, but, of course, that isn’t enough. At the same time, you need to have a deep level of knowledge about the company you invest in to properly calibrate the risks and assess its value creation for more than a few years ahead of time. All that requires you to know a lot about the business, its nature, strategy, and culture which in turn requires tons of work and reading. Deep research is what you put your money on the line for. The rabbit hole is your friend. But the nature of deep research is that time runs away from you. When you decide to go really deep into a potential investment, you dig into the filings, build your model, think a little, read some more, search the web, write down what you think, and suddenly it’s been days, maybe weeks, until you’ve finally reached a conclusion. And then, after that, you have to fight your inner sunk cost bias of not wanting to let go if your conclusion has been incohesive and all your work has been in vain.
When studying businesses, no work is really wasted. The important thing is foremost to do the actual work (while not fooling yourself) and compounded knowledge will eventually do its magic to give you the occasional big insight sometime in the future. However, if deep research is all you do without looking at the entire universe of potential investments, missing the forest for the trees becomes a real risk to your stock picking effort. Meanwhile, if all you do is study compounders off your list of Twitter follows, systematically scan for low price-to-books, or dive into companies off the 52-week low list, you risk becoming Munger’s one-legged man. In order to approach the investing game through the lens of a business owner, you have to not only understand one type of company but the whole spectrum of them. You need to be able to instantly tranche businesses into good, bad, and everything in between. Most importantly, you need mental examples of how seemingly great financials can lead to business disasters (I don’t want to say SVB here, but yes, SVB). Otherwise, how do you even determine your hurdle rate? How do you determine the average height of a population let alone assess what a ‘tall’ or ‘short’ person looks like if you haven’t measured each individual first?
Here, you might say: “But Oliver, averages and shortcuts, that’s what ratios and multiples are for”, and you would be partly right. That’s why screeners are popular. But you need to bring another element to the analysis other than looking at cold-hard figures. For every “ideal” ratio you feed your screener, you lose out on many learning (and investment) opportunities. Take margins. What does the typical gross margin look like for a retailer with ostensible pricing power? 25-30%? Say you fed your screener with a floor of 20% gross margin amongst other filters. Then you missed out on Costco for the company’s entire life. Consequently, you’d disregard the fact that all of Costco’s secret sauces (culture, cost paranoia, purchasing power, and scale efficiencies shared with the customer) in the end originates from the company’s puny 12% gross margin. Or if you take screens for minimum ROIC or FCF, there’s a good chance that Amazon would never have appeared. My favorite example of a faulty screener is one that requires positive working capital, failing to include all those moaty companies that grow using negative cash conversion cycles.
So I don’t use screeners. Instead, I took my idea about stock picking being like book reading (skimming a lot, reading few, and only re-reading the best ones) and I took that idea seriously, creating an exercise in the spirit of Buffett’s early approach that I call:
Once or twice per week I spend an entire day blazing through a bunch of stocks, spending maybe as little as 2 min and as much as 15 min on each. I pick a cluster of companies for each session, either from the same industry/ecosystem or a completely random bunch with different market caps and economics to vary the learning process. I go through them one by one, spending as little time as I need before boredom to maximize the number of stocks I can expose myself to in one session.
When doing a stock sprint, the idea to gain efficiency is to get the most important components of a company in the least amount of time, preferably being fed the components in the same way for each company to develop pattern recognition. ValueLine is an excellent example of that since the more ValueLine pages you glance over, the quicker you will be to glance over the next. You can, of course, pay for ValueLine (and yes, it’s digital now although you can still get the classic print) but there are some other great (and partly free) sources that I currently use instead. Here are three good ones:
- ROIC.AI: An alternative to ValueLine with fundamentals going back 30 years.
- QuickFS: An equally concise source of financials that’s as simple as it should be.
- Koyfin: Great for charting fundamentals against each other.
As I page through businesses, I try to have fun with it by treating it as a game. Sometimes, I look at the financials first without looking at the company name and see if I can guess what kind of business it’s in. Other times, I try to memorize the economics and how they have developed over time, make a note to revisit the company in x time, and then see if I get better at making numbers stick in my mind. Sometimes I visit the company’s website and read its history. Sometimes, I may read just the shareholder letter in the annual report and relate what I read to the numbers in QuickFS. Other times, I study the footnotes in the annual report first since reading those alone can often tell you more about how the company operates than reading an entire case study on the company. The key idea is to switch up how I study companies while anchoring to the financials as displayed in ROIC.AI, QuickFS, or ValueLine and while keeping the study short and succinct.
Each study ignites a series of questions in my mind in order to understand what’s truly going on behind the numbers. In doing so, I like to think of myself as a skeptical journalist. Here are three starting questions I ask every time:
- “What’s in the book? What’s invested in the business and what are the returns?”
- “Does it look like a good business? Is it getting better as it grows?”
- “Does it look cheap? If yes, what am I missing?”
Which may lead to more granular why-why-why-kinda questions like:
- “Why did the company generate huge returns on capital in one year and low single digits in the next?”
- “Did they do any major acquisitions in the meantime?”
- “Does variable RoC mean there’s no durable competitive advantage?”
- “Would taking the average misrepresent reality and why?”
- “Could this variability provide an opportunity in how the market may price the stock?”
Just like skimming a book equals less retention, you have to make an effort during a stock sprint to make things stick. Even though I might only spend a few minutes on each company, I make sure (or at least try) to write every thought I have down, even though these thoughts may later turn out to be ridiculous (or blank) and the company may be way outside of my circle of competence. X% of the stuff I write down is going to be pure speculation to get the brain going and I may never look at the company again. But I might, and that’s why I write it down in real time. That way I can remember my first speculations as I later learn more about the business (either in the passing reading a random article or deliberately). I can couple new inputs with what I wrote down a while ago, sparking a compounding learning curve.
I do this in Roam Research, a bi-directional note-taking system, an indispensable tool to my research process, and in my opinion the best note-taking app ever created. The thing about a bi-directional note-taking system as opposed to a hierarchical one is that it allows you to “rediscover” information because it automatically creates relationships between related pieces of knowledge. Say I made a note to myself on Texas Instruments six months ago reading: “Texas Instruments is a cyclical slow-grower (or non-grower) but maybe has a nice moat since it doesn’t do what every other hyped semiconductor businesses do. The majority of the business is in analog chips, a different part of the value chain with different economics and competitive forces. They spend shockingly little on R&D for a semico.” I then make the word ‘Texas Instruments’ a hyperlink which creates a page for Texas Instruments in my Roam database. Now say that six months later I read a long article that mentions Texas Instruments in the passing which holds a nice piece of information that I save in my notes which might say: “Texas Instrument’s ability to generate huge revenue with minuscule R&D and S&M lies in its customer relationships. Meanwhile, their analog chips are sticky, selling for 50 cents to a couple of bucks per piece with hundreds of thousands of SKUs”. So I save that phrase with the inference that there’s little chance of any competition to take a meaningful piece of their business and that little note (or ‘node’ in Roam terms) may be buried within a bunch of other stuff I saved from the article. And then a month later say I read the annual report, make a bunch of highlights including how the company generates huge margins and short payback period on each fab built together with some wonderful unit economics, and then export those notes to Readwise which then import those to Roam. Now, if I was to look at the Texas Instruments page in the database, all ‘nodes’ concerning the company will automatically be there, excavated from wherever it was buried in the database, allowing me to digest how my thought process and findings about the company has developed over time. This might be confusing stuff but it’s actually simple, automatic, and extremely valuable. Perhaps I will write a longer article in the future showing Roam’s wonders and how I exactly use it.
Once I look at a company during a stock sprint and think “hey, this might be interesting to explore further”, I add it to my research kanban in Notion which is really my ‘annual report reading list’ from where deeper research originates. Before I do deeper research I ask myself the following question: “Can I make any qualified guess about the terminal value of the company?”. This is a nice question to ask since it uncovers:
- Whether the business is predictable.
- Whether I understand its place in its ecosystem.
- Whether I grasp the company’s value creation, not only in the recent past by the proof of its numbers, but also in the long run.
Now if I determine that there’s just an inch of a question mark on the company’s terminal value, I know that I’m intrigued by the business for the sole purpose of learning and expanding my circle of competence, and therefore investing in the company is out of the question unless a miracle happens and I’m truly surprised. Even an average analyst (like myself) will only do great work if they’re interested in what they’re doing with an insatiable curiosity and courage to follow that curiosity. This is also the remedy to sunk cost bias. Sunk cost bias has its way of forcing you to find an insight that isn’t there. If you know beforehand that you study businesses for the sole purpose of learning and expanding your circle of competence with a healthy dose of humility, you can let the occasional big insight come to you without forcing it. And if you decide to leave it, you just head on to the next one.
For more on how I approach deeper research after a company has entered my kanban, read my research process.