My investment method (part 3)

Here are parts 1 and 2.

We’re now in the afternoon session of my typical day, when the wide funnel from the morning narrows into a small set of potential opportunities.

As I wrote in the last part, this is when I continue to kill ideas after digging a bit deeper past the filtering stage, and that’s by design. Good stock picking is unglamorous work with brief moments of excitement at the edges. It’s a loop of reading, triangulating numbers, cross-checking claims, and building context. The spikes of adrenaline — like finding a ridiculously mispriced situation or getting punched in the face by the market — are the moments when you need to be calmest.

It’s counterintuitive because it’s, in a sense, the inverse of how most people move through their careers. They try to avoid the grind to chase thrills. But stock picking is about embracing the grind, with a passion, so the rare thrill has a prepared mind waiting for it.

Something that surprises people is that I rarely spend more than a few days’ work on any one stock, even though my portfolio rarely counts over eight stocks. “You spend two or three days and think you know enough to pile into it?” is one reaction I’ve heard before. The objection is usually shaded in the belief that real conviction requires a month of meetings, expert calls, and a towering 15-tab spreadsheet. In my experience, that path is often more dangerous than beneficial. Many investors strive for certainty and precision in vain, avoiding situations in which information is difficult to obtain (like overlooked stocks). Yet it’s exactly that uncertainty which accompanies the mispriced opportunity. By the time the uncertainty has revolved, the price has already risen.

Truth is, information follows a power law. The first pass over the filings, the balance sheet, the long-run history, and the incentives tells you 90% of what matters that you can know. Believe it or not, there comes a point when negative marginal returns hit your research because the last 10% you try to squeeze out becomes dangerously correlated with sunk cost bias. The longer you spend researching, the more you need the idea to be true.

Fast research wasn’t always my style. I used to spend weeks torturing spreadsheets, layering in scenario trees, and collecting every data point I could find. But I eventually learned how easy it is to mistake activity for insight. A month’s effort on one stock can force a narrative onto the company that the facts within the first hour should have vetoed. The sell side falls into this trap because its product is volume through endless notes, updates, and insider “access” in a tiny subset of the market that gives the analysts tunnel vision. And buy-side analysts at big institutions are subject to their own incentive structures too, where producing more ideas is a hidden job requirement (through competitive pressure), so the marginal investment idea often gets manufactured past the point where it should have been abandoned. If you measure your value by the quantity of ideas, you’ll invent reasons to keep digging.

I find that speed in research is very important because it acts as discipline against waste. If you spend an entire month per idea, you’ll only look closely at a dozen or so prospects a year and perhaps some 500 in a full career. That sounds like a lot until you realize how minuscule a slice that is of the opportunity set. And a huge chunk of that research time will be wasted, making the useful work an even tinier fraction. The market rewards prepared opportunism. So you want to turn as many stones as possible without letting the process turn sloppy.

I move quickly through my research because I intentionally read quickly and try to read in the right order. But what I’m not doing is skipping. An annual report might run anywhere from 80 to 400 pages. I flip every page. You don’t want to read to harvest quotes, but to reconstruct the business in your mind and investigate it like a journalist would, sniffing out the truth in a balanced way. By the time a stock has gone past the filtering stage, I’ll already know the gist of the investment case, whether it’s a hidden asset, a blurred segment, or a mispriced earnings stream, otherwise I wouldn’t have written down the name in the first place. So the whole point of my afternoon work is to sprint through the right questions, slow down only where those answers deserve it, and stop when the odds don’t line up.

Another point of my research approach is that edge comes from doing the hard (or rather boring) things that actually change your odds, like reading primary documents front to back, adjusting accounting numbers to reflect economic reality, mapping incentives, and learning the long arc of the business rather than last quarter’s noise. This requires skills in accounting, valuation, and a little psychology, but not nearly as much skill as people think.

What’s important in stock picking is judgment as opposed to informational access. Fifty years ago, learning about a smallcap might have meant a trip to the library, sending away for a paper filing, and hoping what showed up wasn’t stale. Today, almost everything is at your fingertips. Sure, the notion that “anyone can know everything” may be exaggerated, but it’s directionally right.

Judgment begins with intellectual honesty. That means resisting “outside-in” narratives at all costs. An outside-in narrative means starting with someone else’s story, like an excited thread on X, a glossy presentation deck, a confident note, or even a conversation with management. These inputs will try to influence you, win your attention, and ingrain themselves in your process, which is their nature because they’re pushing to be as important to you as possible. If you shortcut the work and rely on secondary sources (I’ll include LLMs here, which tend to give you what you want to hear), you’ll miss the fact that many of the investment theses you read out there tend toward lopsidedness. How many times have you read a glowing pitch for a stock, only to poke at it yourself and found that the author conveniently left out a fatal risk, either out of incompetence or sinister reasons?

There’s a whole marketing aspect to published investment research. If you produce research for subscribers, your PM, or even your friends, you’re subconsciously pushed to sound more certain than you are. And markets have long feedback loops. You can talk with conviction today, forget the miss two years later, and the people listening to you will forget even sooner. It’s cheap to sound authoritative when the bill comes due in slow motion.

The goal in investing is to maintain intellectual purity, because the human brain craves coherence. Once an idea pops into the mind, the mind instinctively fills gaps to protect that idea. My rule is therefore simple: filings first, always. If someone sends me an idea, I jot the name down and look at the filings before reading the rest of their message. Likewise, I stay far away from investor presentations before getting to the end of my process, when I can use them as a quick recap of things that are already in the filings. And I always look at the balance sheet, cash flow statement, and income statement (in that order) first, before reading the full interim or annual report.

The other pillar of judgment is being aware of how your own head tries to trick you in the research process. Here are the most important traps to actively guard against:

• Sunk cost bias. We already talked about this: the more time and energy you pour into an idea, the harder it becomes to walk away. Short, intense research sprints keep this in check because decisions arrive before your mind develops ownership over the work.
• Framing and anchoring. The earliest data point you encounter — the first valuation multiple you see, the first narrative you hear, the first chart you glance at — becomes the frame around everything that follows. Anchors stick even when we know they’re arbitrary.
• Dunning-Kruger and chauffeur knowledge. Superficial familiarity can easily masquerade as confidence. I write while I read specifically to expose gaps in my understanding.
• Institutional imperative. This means doing what looks “proper” because others do it. In stock picking, convention can be corrosive. Independence of mind and process is about ignoring performative habits that don’t move the odds.
• Path dependence. Be aware of conclusions that require many things to go right at once. If the investment case only works when a long chain of contingencies lines up, your conviction should be low even if each link looks probable in isolation.
• Narrative fallacy. Once a story exists, and it’s a good story, risks tend to get edited out. Make it a habit to write a “kill list” of ways it can go wrong before writing the bull case.

At this point, you might reasonably ask what I’m looking for in the deep dive that I didn’t already see in the filtering stage. And the answer isn’t a single metric or pattern. I’ll rather call it “coherence”. Most ideas fail the coherence test. The unit economics look fine until you notice that customer acquisition is subsidized by cheap financing that’s going away. Or the price looks attractive until you adjust for recurring restructuring and stock comp that management treats as ornamental. A good idea, by contrast, keeps aligning as you widen the circle: the shareholder communication matches the cash flows, or the proxy explains insider behavior. In other words, in a good deep dive, you wanna move from “it looks cheap” to “it makes sense.”

With that, here’s my process for any deep dive I do into a stock:

First, a bit on my note-taking. It’s structured and quite consistent between companies, which makes it easier for me to compare notes between companies and compartmentalize the information in my mind. So I don’t keep one long, amorphous note. Instead, I have nine major buckets in my single company note that I drop information into. Each bucket is literally a section header with nested multi-level bullet points under it. As I read more, I start linking and nesting information next to and under each other. (I keep all my notes in Obsidian, btw.)

The nine buckets are:

1. What: This is where I put everything about the business, what it does, how it makes money, its history, footprint, stakeholders, scale, and so forth.
2. Growth: Notes about recent growth and the reasons, guidances, and underlying drivers. This is also where I put notes about industry structure, market size, market growth, and peers/competitors. I want to know how the company grows and at whose expense.
3. Profitability: Notes about current and historical margins, and the moving pieces behind them. I disaggregate what’s happening above the gross profit line, what’s going on in SG&A, what’s driving cash taxes, the cost of debt, and more. I keep separate nested bullet points on each driver rather than one blended margin line that hides everything.
4. Reinvestment: Anything related to capital allocation in terms of capex, working capital, acquisitions, buybacks, dividends, debt paydowns, and more. I care less about what management slogans say about capital allocation and more about what has actually happened to cash over time.
5. Risk: Anything related to leverage, customer/supplier concentration, regulatory risk, tech obsolescence, corporate culture, key-person risk, barriers to entry and exit, macro, and more goes in here. The “kill list” of ways I can be wrong lives in this bucket.
6. Non-operating assets: Anything that’s not essential to the operating business, like excess cash, equity stakes, real estate, idle assets, and more. Accounting conventions can throw these off significantly, so I try to mark what I can to market. I typically focus on land and buildings. Opportunities often arise from opaque balance sheets misunderstood by the market.
7. Valuation: The bucket where I put my thoughts about three valuation levels: liquidation value, asset reproduction value, and earnings power value. I also put notes on potential catalysts and my running answer to “why the opportunity?” Rather than aiming for precision, I want ranges and sanity checks. (I’ll delve more into how I approach valuation in part 4.)
8. Alignment: Anything related to ownership structure, insiders, compensation, incentives, related-party deals, governance, and more goes in here.
9. Accounting: Here I put notes on idiosyncratic accounting quirks specific to the firm and its industry, and anything that distorts the economic reality of the business.

Here’s the typical order I read things in:

I always start with the last interim report/10-Q to dig into the balance sheet. I mentally divide every asset into operating vs non-operating assets and look at how the whole thing is funded, gauging its fungibility and safety. The reason I start with the latest interim report is simply to get the freshest balance sheet, the right number for shares outstanding, and to get up-to-date on any recent events like new debt and such. Other than that, no long-term investor is going to find anything in the last quarter that’s particularly useful. It’s not necessary to check in on the revenue and earnings figures of the stocks you own more than once a year.

Next, I read the full latest annual report or 10-K. (If I have the option of both, I stick to the 10-K to get the information as raw as possible, without the gloss.) The 10-K is where I spend the most time because the 10-K is the most useful document about a company there is.

As I read the report, there are two points that I focus on, which I think differ from how many investors approach it.

1. I think in terms of the enterprise, not just equity. Investors who fixate on the equity slice end up staring at EPS lines and forward guidance only to lose the forest for the trees. That approach can make you forget that value ultimately sits at the business level, not in whatever slice of the cap structure you happen to own. I think about and try to analyze the relationships and conflicts of interest between each claimholder in the structure. For instance, secured lenders want asset coverage, bondholders care about covenants and solvency, preferred shareholders want stability, and common shareholders want upside. When you lay all the pieces out, you get a far truer picture of how the business functions and why management makes the decisions it does. I also attempt to put a market value on each claim to the business. That means, in calculating enterprise value, I use not only the face value of debt but its market value, the market value of preferred stock, any off–balance sheet obligations, non-controlling interests, pension deficits, and whatever else that constitutes an economic claim on the firm. Enterprise value should reflect the true cost of acquiring the business and settling every claimholder, not the accounting shortcut that often gets passed around.
2. I don’t appraise the business purely as a going concern. While everyone cares about earnings power, margins, and growth, that’s only half of the equation in terms of how a business can generate value for its shareholders. The other half is resource conversion: selling divisions, liquidating assets, refinancing intelligently, extracting value from non-operating assets, and so forth. Most investors tend to completely ignore this part of the coin, treating the business as a perpetual-motion earnings machine and forgetting that resources can be redeployed, sold, written down, or revalued in ways that materially affect per-share value. In fact, some of the best investments come from companies where resource conversion — and not necessarily earnings growth — is the dominant story. So I read the report with two lenses: earnings power and resource conversion.

After reading the latest 10-K, I like to jump straight to the oldest 10-K that I can find for the company. Hardly anyone does this, which is why it’s a useful thing to do. Reading the newest and oldest reports back-to-back is a fast shortcut to understanding how the business and its industry have evolved over a long period of time. You see how the revenue mix changed, which segments disappeared, how margins and returns moved, how leverage changed, and how management’s language shifted, or didn’t shift, over time.

If there’s an S-1 filing, other IPO prospectus, or a spin-off document, I read that too. Though long and tedious, they tend to contain the best historical and structural context you’ll get in one place.

I’ll typically also read every shareholder letter from the earliest annual report to the present. Many shareholder letters are just a couple of pages, regurgitating last year’s talking points or offering boilerplate macro filler, so reading through them shouldn’t take long. The best ones are when management is trying to communicate honestly rather than perform. You can tell a lot about a management team by what they choose to emphasize, what they ignore, and how honestly they explain the decisions that didn’t work. I pair the shareholder letters with quickly reading through the earnings transcripts.

I skim a little more of the past reports for the years when an inflection like a CEO change, capital raise, special dividend, big acquisition, and so forth has occurred.

(A quick tip: when trying to match up what management in the past said it was gonna do with what actually happened, try not to compare the point-to-point income statements, but the balance sheets, which are much harder to fiddle with.)

I always read the company announcements — PRs, 8-Ks, 6-Ks — for the past twelve months. These updates often look trivial, but form the real-time breadcrumb trail of management’s priorities. And if you care about resource conversion or special situations, this is where the action shows up first, such as asset sales, recapitalizations, insider moves, credit amendments, spin-offs, quiet shutdowns of failing segments, and so forth. The income statement won’t tell you these things in time, but the announcements will.

Next, I move to the latest proxy statement (something like the DEF14A for US-listed companies). This is an important document that many skip. It’s where you learn who controls the company, who influences the agenda, and how management is compensated. Incentives are often the missing key in understanding what a company will do rather than what it says it’ll do. A beautifully written shareholder letter means little if the CEO’s compensation is tied to metrics that encourage short-term financial engineering. A CEO with 1% ownership and a compensation plan tied to adjusted EBITDA will behave very differently from a founder with 40% of the stock and no cash bonus.

Then, and only then, do I sometimes crack open the company’s investor presentation. This may surprise you because investor presentations are made for the “benefit” of shareholders, designed to clarify the business, give you a tidy picture of the value prop, and help you understand the company and its industry quickly. But that’s why this is the most dangerous document to read first. Investor presentations are marketing docs. Every chart slopes up and to the right, every data point has a satisfying narrative, and every claim is dressed in a kind of corporate earnestness designed to disarm your skepticism. The irony is that the investor presentation often contains a lot of useful information, helping accelerate your understanding of the industry dynamics, packed into a 10-minute deck. That’s why I save it for last.

By the end of reading, my company note has a bunch of messy points scattered across the nine buckets I described earlier. That’s when I sit down and clean it up, nesting related points and connecting dots.

So now, I’ve moved through the primary material, after which my research becomes more random. This is where I start following threads. If a footnote mentions a concentrated customer, I’ll pull filings from that customer. If the company operates in a niche I don’t fully understand, I’ll read a trade journal or dig up whatever I can on the industry. I’ll skim or read competitor 10-Ks too. Sometimes this rabbit hole lasts a few hours, and other times it lasts a day or two. You can dig up far more than you think if you’re willing to chase your way through odd sources, like counting product reviews, tracking distributor inventory, reading federal grant databases, tracking large customers through their earnings calls, scanning regulatory databases, searching Glassdoor, and so forth.

This is also where scuttlebutt can help, but don’t imagine I’m calling 20 ex-employees or spending weeks talking to experts via Tegus or to competitors (as I’d rather spend time blazing through their reports). Those activities can be useful, but they’re slow, biased, and easy to overinterpret. Outside of the filings, your time and energy are best spent trying to figure out one single question: whether customers are happy and why. This is something you’ll never get a straight answer to from management, and GAAP won’t tell you either.

I treat talking to management teams the same as I do investor presentations. It’s fine at the end of the process, once I understand the business on my own terms. That’s because management, no matter how honest, is still an agent, and you as a shareholder are the principal. Buffett compares this business to investigative journalism for a reason. A journalist investigating a politician wouldn’t start by asking the politician for the truth. Instead, they’d pore through documents, public records, and third-party accounts before approaching the subject. Stock picking works the same way. Also, you don’t assess a horse by talking to it. You assess it by watching it run, race after race, across multiple surfaces, in different conditions.

You should be 80% toward a buy decision before you reach out to management or IR. Otherwise, you won’t stand a chance against the biases that creep in, such as liking bias, selection bias, authority bias, commitment and consistency, framing, confirmation bias, the narrative fallacy, or the focusing effect. If you speak to charismatic people too early, you let their story set the frame. Reaching out to them does, however, tell you something subtle but important: how the company allocates its attention. Some management teams are nearly impossible to reach because they’re running a focused operation. Others will hop on a call with any tiny investor, anytime, because they’re desperately managing sentiment. I prefer the former, even if it means I’ll have an unanswered question.

Throughout my full research process, it’s important to me that I continuously write to resist passively sitting back and consuming content. It’s hard to make things stick just by reading. The mind happily nods along to things it doesn’t actually understand. But when you force yourself to write — to put the business into your own words, to map the logic, and to reconcile contradictions — you discover quickly where your understanding is shallow. Writing also exposes the cracks. Whenever I’m stuck writing, I know that I need to collect more pieces of the puzzle.

At this point, you’re probably wondering: “When does he actually model a damn thing?”

I’m not one of those value investors who swear by back-of-the-envelope calculations for everything. Spreadsheets are awesome, and I believe in using them. The answer is I do build company models, but it’s for the purpose of helping me pound the numbers into my brain. I enter data manually into my model on the go as I read the reports, focusing mostly on disaggregating the financials into segments and geographies while also tracking KPIs and adding whatever unique data sources I can find. Karelia is an example, where you see I mixed numbers and commentary to gauge organic volume growth in the cigarette business.

There are companies that work better in a spreadsheet than others, like restaurants or anything with nice disclosure on the unit economics. And then there are the levered cases, where modeling helps you understand how the company’s capital structure and capital allocation interact over time. Other businesses need no modeling at all.

I also use this spreadsheet I’ve built to adjust a company’s financials to get closer to its economic reality. Financial statements are a starting point in valuation, not the truth. What matters in the end is getting to the right figure of owner earnings. Some of the adjustments I make are, for example, capitalizing R&D expenses (which depress earnings and understate invested capital), adding back write-offs and impairments to the invested capital (to penalize past poor capital allocation decisions), or smoothing or reclassifying lumpy “other expenses” over a longer period. I do the same thing for the company’s competitors. It’s about getting to a place where the economics aren’t blurred by accounting makeup, which allows you to compare appropriately across companies and value the business honestly. Accounting is a language filled with tripwires designed either to trick you into believing a business is better than it is or to blind you to the needle in the haystack.

Now, this would be a natural moment to segue into valuation — how I translate all of this into estimates of intrinsic value, why I don’t think of valuation as a single point but a distribution, how I think about opportunity cost, how the margin of safety underpins it all, and why most people dramatically overcomplicate valuation while simultaneously getting the research process wrong.

But since this ties neatly into how I finally put money to work, how I size positions, how my ideas fit into my two investment buckets (special sits and generals), and when/why I sell stocks, I’ll save the valuation aspect for the next part.

Part 4 is coming this week, so keep an eye on your inbox. I hope you enjoyed this one, and if you know someone else who might too, feel free to share it!

Cordially,
Oliver Sung