How Do You Research Stocks? The Full Process, Stage by Stage

You research a stock by working through eight stages in order: understand what the business sells and how it makes money, read the filings that matter (the 10-K, the latest 10-Q, the proxy), work through the numbers (growth, margins, cash flow, balance sheet, dilution, returns on capital), test whether the competitive position is real, judge management and their incentives, put valuation in context, build the bull and bear case, then write the thesis down with the triggers that would change your mind. On a company you don't already know, that's six to fifteen hours of work, not an afternoon. This article is the full process, stage by stage. It is educational only and not financial advice or a recommendation about any security.

How do you research stocks?

You research stocks by running a repeatable, ordered process: business model first, then primary filings, then the financial statements, then the moat, then management and incentives, then valuation, then both sides of the argument, then a written thesis with tracking triggers. The order matters. Numbers whose source you don't understand are just decoration.

If you only have five minutes, don't start here. Start with the triage version: read a stock in five minutes is a six-question filter that tells you whether a company deserves any more of your attention. Most names should fail it. This article is what you do with the survivors: the full workup you run before real money moves.

The stock research process at a glance

Stage	What you're answering	Where to find it
1. The business	What does it sell, to whom, and what does one unit earn?	10-K Item 1, investor day decks, the company's own pricing pages
2. The filings	What is management legally required to tell me?	10-K (Items 1, 1A, 7, 8), latest 10-Q, DEF 14A proxy
3. The numbers	Is the growth real, profitable, cash-backed, safely financed?	Income statement, cash flow statement, balance sheet, footnotes
4. The moat	Why can't a well-funded competitor take this away?	Gross margin trend, pricing history, churn, peer filings
5. Management	Are they paid to do the thing that makes me money?	Proxy statement, prior-year earnings calls, buyback and M&A record
6. Valuation	What does today's price already assume?	Peer multiples, the company's own 5 to 10 year multiple range
7. Both cases	What has to go right, and what breaks it?	Everything above, argued in two directions
8. The written thesis	What would tell me I'm wrong?	Your own notes, dated, with numbered triggers

Stage 1: Understand the business before you touch a number

Write one paragraph, in your own words, with no jargon, explaining how the company turns effort into cash. What's the product, who signs the check, how often, and what does the company keep after the direct cost of delivery? Then go a level deeper into unit economics: what does it cost to acquire a customer, how long do they stay, how much gross profit do they generate over that life? For a retailer: what does one store cost to open and what does it earn once mature? For a subscription business: what's the gross margin on one seat?

The test is whether you can explain the company to someone who's never heard of it and answer their follow-ups without looking anything up. Then write down the two or three variables that actually drive the model. Usually it's a short list: volume, price, cost per unit.

Stage 2: Read the filings that matter

Three documents carry most of the weight, and each answers a different question.

• The 10-K. Item 1 (Business) gives you segments, customers, and competition in plain English. Item 1A (Risk Factors) is where management is legally obliged to name what could go wrong. Read Item 1A against last year's version: anything that got longer, newer, or more specific is a signal in itself. Item 7 (MD&A) is management explaining results in their own words. Item 8 and the footnotes are where the accounting truth lives. Our guide on how to read a 10-K covers the section-by-section method.
• The latest 10-Q. The 10-K tells you what the business was. The most recent quarter tells you what it is. Compare them: is the trend from the annual report intact, or did something turn?
• The proxy statement (DEF 14A). Almost nobody reads it, which is exactly why it's valuable. It's where you learn what management is actually paid to achieve.

All of this is free on the SEC's EDGAR database. If you're researching a private company or a small business (an acquisition target, a supplier, a competitor you're sizing up), none of it exists and you may be working from raw bookkeeping exports. In that case it's worth turning those exports into readable financial statements first, because you can't analyze what you can't line up year over year.

Stage 3: The numbers, in the order that matters

Pull five to ten years of financials if the company has them. Three years isn't enough to see a cycle. Then work this sequence:

1. Revenue growth and its quality. Growth alone means nothing. Break it into price, volume, acquisitions, and currency. A company growing 20% by buying revenue at expensive multiples is a different animal from one growing 20% organically. Check whether growth is decelerating and whether backlog or deferred revenue is keeping pace with reported revenue.
2. Gross and operating margin trend. Direction beats level. A 40% gross margin heading to 45% is a better story than a 60% margin heading to 52%. Falling gross margin alongside rising revenue usually means the company is buying growth with price.
3. Cash flow versus reported earnings. Compare cumulative net income to cumulative free cash flow over five years. If earnings consistently exceed cash flow, find out why: rising receivables, capitalized costs, aggressive revenue recognition. Persistent gaps are among the loudest risk flags there are.
4. Balance sheet and debt maturities. Don't stop at total debt. Find the maturity schedule in the footnotes. Debt due in seven years at a fixed 3% is a footnote. The same amount due in eleven months, floating, is the whole story.
5. Share count. Track diluted shares outstanding over five years. If the count keeps rising, per-share growth is being quietly taxed by dilution, and buybacks that only offset stock compensation are not returns to you.
6. Returns on capital. Return on invested capital, tracked over years, tells you whether the company creates value when it reinvests. A business earning high returns that can redeploy cash at those returns compounds. One earning less than its cost of capital destroys value every time it grows.

Stage 4: Test whether the moat is real

Every management team claims a competitive advantage. Your job is to falsify the claim. A real moat leaves fingerprints in the numbers: stable or rising gross margins across a full cycle, the ability to raise prices without losing volume, low churn, and returns on capital that stay high for years while competitors try to compete them away. If margins are eroding and the company still calls itself the category leader, believe the margins.

Then ask the falsification question directly. If a competitor with a billion dollars and five years decided to take this business, what exactly would stop them? Switching costs, network effects, scale in cost, a regulatory license, a brand that lets the company charge more for a physically identical product. If the honest answer is "nothing structural, they're just executing well," that isn't a moat. It's a head start, and head starts close.

Stage 5: Management and incentives

Open the proxy and find out how the CEO is paid. If the bonus keys off revenue growth or adjusted EBITDA, expect acquisitions and add-backs. If it keys off return on invested capital or per-share metrics over multi-year periods, expect more discipline. Incentives predict behavior far better than the mission statement does.

Then audit the capital allocation record. Over five years, where did the cash go: reinvestment, acquisitions, buybacks, dividends, debt paydown? Were buybacks done when the stock was cheap or when it was expensive? Did the acquisitions earn their cost? Finally, do the thing almost nobody does: pull an earnings call transcript from a year or two ago, write down what management promised, and check what happened. A team that quietly changes the subject each year has told you something.

Stage 6: Valuation in context

A multiple in isolation is meaningless. Put it in three contexts. First, against close peers doing roughly the same thing, adjusted for growth and margins. Second, against the company's own multiple range over five to ten years, so you know whether today is expensive relative to its own history. Third, and most useful: reverse the question. Instead of asking what the company is worth, ask what growth and margin the current price implies. If the price only works with 25% growth sustained for a decade, you now know exactly what you're betting on.

A cheap multiple is often a trap. Cyclicals look cheapest at the peak, when earnings are highest and about to fall. A structurally declining business gets cheaper every year on the way down and stays a bad investment the whole time. A low multiple is usually the market pricing something real that you haven't found yet. Find out what it is, then decide whether you disagree.

What should I look for before buying a stock?

Look for a business you can explain, growth that's cash-backed rather than accounting-backed, a balance sheet that survives a bad year, an advantage visible in the margins, management paid to do the thing that benefits you, a valuation that doesn't already require perfection, and a written bear case you genuinely respect. If any of those is missing, you're speculating.

The disqualifiers matter as much as the positives: debt maturing soon that cash flow can't cover, earnings that never turn into cash, one customer making up a large share of revenue, serial "one-time" charges, a rising share count with no matching growth. Any one of these can end a thesis on its own, and it's far cheaper to find them now than after you've bought.

Stage 7: Build both cases, then decide what would change your mind

Now argue both sides properly. The bull case and bear case should each be specific, quantified, and falsifiable. The bear case especially has to be one you'd genuinely worry about, not a straw man like "the valuation is high." If you can't state it as well as its strongest advocate would, you don't understand the company well enough to own it.

Then write down, in advance, what would make you change your mind. Not vague discomfort: specific conditions. Gross margin below a level for two consecutive quarters. Loss of the largest customer. Net debt crossing a threshold. The founder leaving. Pre-committing while you're calm is the only reliable defense against rationalizing later, when you hold the position and the news is bad.

Stage 8: Write the thesis down and set the tracking triggers

This is the stage almost everyone skips, and it's the one that compounds. Write a one-page memo, dated: what the business does, why it should do well, what has to be true, what would break it, what the price implies, and the specific numbers you'll check each quarter. Then actually check them.

Two things happen. You catch thesis drift early, when the reasons you originally bought quietly stop being true. And a file of dated memos becomes the only honest record of your own thinking. Memory rewrites itself in favor of whatever happened. Your notes don't.

How long should stock research take?

For a company you don't know, budget six to fifteen hours across more than one sitting: about an hour on the business model, three to five on filings, two to four on the financials, one to two on management and the proxy, one to two on valuation, and an hour writing it up. Maintenance is then roughly an hour per quarter per position.

Spreading it over days isn't inefficiency. It's how you catch what you missed on the first pass, once your brain has stopped trying to reach a conclusion. The point isn't the hour count. It's that fast enthusiasm is the most expensive way to buy anything.

What is the most important part of stock research?

Understanding the business model well enough to know what actually drives it. Every other stage depends on it. Valuation without it is arithmetic on numbers you can't interpret, and the bear case you write will miss the risk that matters, because you won't know which variable the whole company hangs on.

The close runner-up is writing the thesis down with triggers. Research that lives only in your head can be revised silently to fit whatever you already want to do, which means it stops being research.

Where AI compresses this, and where it doesn't

The real bottleneck is gathering and structuring raw material: pulling ten years of financials into one view, extracting company-specific risks from a hundred pages of risk factors, lining a company up against peers, and drafting both sides of the argument so you have something to react to instead of a blank page. That work is mechanical and slow. Investables.ai compresses it: enter a ticker and you get a structured research card with a thesis, both cases, key metrics, comparables, and risk flags, which is your stage 1 to 3 raw material assembled in seconds rather than an evening. Our roundup of stock analysis tools covers where the rest of the stack fits.

What doesn't compress is judgment. No model can tell you whether a moat holds for a decade, whether this management team is honest, whether you'd sit through a 40% drawdown, or whether the risk the filing describes is the one that actually bites. AI gets you to the starting line faster with better material. The decision stays yours. Nothing a research card produces is financial advice, and past performance does not guarantee future results.

The output of good research isn't a decision to buy. It's a written statement of what you believe, what it depends on, and what would prove you wrong.

The bottom line

Serious stock research is a process, not a mood. Understand the business, read the filings, interrogate the numbers, test the moat, check the incentives, put valuation in context, argue both sides, and write it down with triggers. The five-minute checklist decides what deserves this treatment. This is the treatment. Skip stage eight and you'll do all the work and still forget why you did it.

Get started