You found a company that looks interesting. You read the latest earnings recap, skimmed a couple of analyst takes, watched a video, and bookmarked an investor letter you'll "get to later." Two weeks on, you can't reconstruct what you actually learned — and when the stock drops 15%, you have no written baseline to check your panic against. The reading happened. The research didn't.
The fix isn't reading more. It's having a research system: a repeatable way to capture what you learn, organize it so you can find it, and revisit it on a schedule. By the end of this guide you'll be able to set one up from scratch, know exactly what to capture for each company, and turn a pile of disconnected notes into a decision you can defend.
TL;DR
- A research system has three jobs: capture (get information out of your head and browser), organize (so you can find it later), and review (so old research stays alive).
- Give every company one durable note — a living dossier you update over time — instead of scattering findings across bookmarks, tabs, and chats.
- Capture in your own words, not copied passages. The act of rephrasing is the research; a highlight you didn't summarize is a highlight you didn't understand.
- Separate facts (what the company reported) from interpretation (what you think it means). Mixing them is how you fool yourself later.
- The payoff is retrieval: when a decision lands in front of you, you want last year's reasoning in ten seconds, not a memory you've unconsciously rewritten.
Why "just reading" fails
Start from something you already know: studying for an exam. You learned long ago that re-reading a textbook feels productive but barely moves the needle, while writing things in your own words and testing yourself actually sticks. Investment research is the same skill wearing a suit. Passive reading produces familiarity — "oh yes, I've heard of this company" — which your brain mistakes for knowledge.
The second failure is fragmentation. Your findings end up in five places: browser bookmarks, a notes app, a spreadsheet, a chat with a friend, and your memory. None of them talk to each other, so when you need the full picture you can't assemble it. The information exists; it's just not retrievable, and research you can't retrieve is research you don't have.
A system fixes both: it forces active capture (you write, therefore you understand) and it centralizes (one place per company), so the work compounds instead of evaporating.
The three jobs of a research system
Every workable system, however simple, does three things. Keep them distinct in your head — most people are strong at one and neglect the other two.
| Job | The question it answers | Failure mode if neglected |
|---|---|---|
| Capture | "What did I just learn, in my words?" | Bookmarks you never reopen; passive familiarity |
| Organize | "Where does this live so I find it later?" | Scattered fragments; can't assemble the picture |
| Review | "Does what I believed still hold?" | Stale theses; reacting to price instead of facts |
The rest of this guide is one section per job, plus a worked example tying them together.
Job 1 — Capture: the company dossier
The core unit of an investing research system is one note per company — a living document you call a dossier. Not one note per article you read; one note per company you're studying. Everything you learn about that business flows into the same place and accumulates over time.
A useful dossier has a stable skeleton so you fill the same buckets for every company and can compare them later:
The company dossier template - What they do — in one paragraph, plain language, no jargon. If you can't, you don't understand it yet. - How they make money — the actual revenue lines, and which one matters most. - The numbers — a handful of figures you track over time (revenue growth, margins, debt, cash). - The bull case — why this could work out. - The bear case — why it could fail. Write this as honestly as the bull case. - Open questions — what you don't yet know and need to find out. - Log — dated entries as new information arrives.
The single most important rule of capture: write in your own words. When you read a 10-K passage about "deteriorating unit economics in the core segment," your job is to write down what that means for this business, not to paste the sentence. The rephrasing is where the understanding happens. A highlight you copied verbatim is a decision you've deferred, not made. (This is the same principle behind [progressive summarization](../ai-notetaking/what-is-progressive-summarization.md): you distill in layers until what's left is genuinely yours.)
Try this: open a note for one company you're curious about and fill only the first bucket — "what they do" in one paragraph, no jargon. If you stall, that stall is your first research finding.
Job 2 — Organize: facts vs. interpretation
Capture gives you raw material; organization keeps it honest. The discipline that matters most here is separating two things that feel identical in the moment but age very differently:
- A fact is something the company reported or that is verifiably true: "Q3 revenue was $1.2B, up 8% year over year."
- An interpretation is your judgment about what a fact means: "8% growth is decelerating and management is downplaying it."
Facts are stable; interpretations are yours and can be wrong. If you blend them in the same sentence, a year later you'll read your own interpretation and remember it as fact — and you'll never notice you were wrong, because you've erased the line between what happened and what you guessed.
| Statement | Type | Why the distinction matters |
|---|---|---|
| "Gross margin fell from 42% to 38%." | Fact | Checkable; either true or not |
| "Margins are collapsing and won't recover." | Interpretation | A prediction you should later grade |
| "Management blamed input costs." | Fact (they said it) | What they claimed, not whether it's true |
| "That excuse won't hold past next year." | Interpretation | Your bet — flag it as one |
Practically: keep facts in the dossier's "numbers" and "what they do" sections, and keep interpretations in the bull/bear cases, clearly marked as your view. When you organize this way, your research connects to the rest of your process — your interpretations become the raw material for [an investment thesis](what-is-an-investment-thesis.md), and the moment you act on one, it belongs in your [decision journal](the-investing-decision-journal.md).
Job 3 — Review: keeping research alive
Research has a shelf life. A dossier you wrote in January is a snapshot of what you believed in January, and the business has moved since. Review is the scheduled act of re-reading your own research and checking it against reality. Without it, you end up holding a thesis that quietly stopped being true.
A light cadence is enough; a heavy one you'll abandon:
| Trigger | What to do |
|---|---|
| Quarterly earnings | Update the numbers; ask whether the bull/bear balance changed. |
| Big price move (±20%) | Re-read the dossier before reacting. Did a fact change, or just the price? |
| A bear-case item comes true | Mark it. Two or three confirmed bear points is a real signal. |
| Periodic sweep (e.g. yearly) | Re-read every dossier; archive companies you no longer follow. |
The "before you react" rule is the highest-leverage habit in the whole system. Price moves trigger emotion; your January self, writing calmly with no money on the line, is often the more reliable analyst. Reviewing first lets that calmer version overrule the panicked one. Staying inside what you actually understand is its own protection here — see [your circle of competence](what-is-your-circle-of-competence.md).
A worked example, end to end
Suppose you're studying a fictional company, NorthCup Coffee.
- Capture. You open one dossier. "What they do": sells subscription coffee direct to consumers; also wholesales to offices. You read their annual report and, in your own words, log: revenue up 8%, but the office segment — their profitable one — shrank.
- Organize. You file the 8% and the segment numbers as facts. In the bear case you write your interpretation: "Growth is propped up by the low-margin subscription side; the profitable segment is shrinking." You add to open questions: "Is office decline temporary (return-to-office) or structural?"
- Review. Next quarter, office revenue falls again. You re-read your dossier, see your own flagged bear point come true, and — because you wrote it down before it happened — you trust it instead of explaining it away. That's a researched decision, not a reaction.
Notice what made this work: everything lived in one place, you wrote in your own words, you kept facts and judgments separate, and you re-read on a trigger. No special tools — just the three jobs, done consistently.
Common mistakes
- Collecting instead of processing. Forty bookmarks and zero summaries is a reading list, not research. If you didn't rephrase it, you didn't research it. (This is the [collector's fallacy](../ai-notetaking/what-is-the-collectors-fallacy.md) applied to investing.)
- One note per article. Scatters a company's story across a dozen files. One dossier per company, always.
- Blending facts and opinions. The fastest way to fool your future self. Mark interpretations as yours.
- Only a bull case. A dossier with no honest bear case is marketing, not research. Write the bear case as hard as the bull case.
- Capturing but never reviewing. Research that's never revisited goes stale silently. The review is where the value compounds.
- A system too heavy to maintain. A plain dossier you actually update beats an elaborate one you abandon in a month.
Summary + next step
A research system turns scattered reading into decisions you can defend. Do three things: capture each company's story in one living dossier, in your own words; organize it by keeping facts separate from your interpretations; and review on simple triggers so old research stays alive. The whole thing fits in a single notes app — each company a note, your interpretations searchable when a decision lands. Keep your dossiers in JustJot.ai, where [semantic search](../ai-notetaking/what-is-semantic-search.md) lets you ask "companies where the profitable segment was shrinking" and surface every dossier that fits, regardless of the exact words you used.
Next step: pick one company, open one dossier, and fill just the "what they do" paragraph today. Then, when you're ready to commit capital, turn that dossier into [a one-paragraph investment thesis](what-is-an-investment-thesis.md) — the argument your research exists to support.