How to research a stock before buying it

Research is a file, not a vibe

Buying a stock without a file is how people confuse familiarity with knowledge. You have seen the logo. You like the product. A friend owns it. None of those facts describe revenue quality, the maturity of the debt, or what happens to margins if a competitor cuts price. FINRA's overview of stocks is a reminder that you are buying a business that can fail, stall, or thrive independently of your opinion of the brand. Research is the work of making that independence visible before money moves.

A usable file has a thesis sentence, a handful of numbers you will update, a list of risks that would kill the thesis, and a note on how the position would change the portfolio you already have. It does not need to look like a sell-side report. It needs to be rereadable in six months when the price is lower and your memory is generous. This article walks through the sections of that file. It is educational, not a recommendation to transact in any company, and StockLift does not execute the purchase if you later decide to make one.

Start with how the company makes money

Revenue is the least glamorous number in a pitch deck and the most important sentence in a research file. What is sold, to whom, how often, and in which currency? A business that bills a few large customers on project work is not the same as a business that collects small amounts from millions of users under contract. Mix matters because it tells you what can disappear in a single meeting. If you cannot explain the mix without looking it up, you are not ready to interpret the growth rate.

Read the revenue discussion in the annual report, not only the highlight reel in an earnings slide. Segment notes, geographic splits, and concentration disclosures are where the story either holds or quietly falls apart. A company that grew because one region or one customer accelerated is not automatically fragile, but it is a different research object than a company whose growth is broad. Write the mix in your own words. If the paragraph requires jargon you cannot unpack, the file is not done.

Margins and free cash flow show what growth is worth

Gross margin, operating margin, and free cash flow are how you see whether scale is helping or whether the company is buying growth with giveaways. Expanding revenue with collapsing margin is a story, but it is not automatically a good one. Free cash flow — cash from operations minus the capital spending required to maintain and grow the business — is the bridge between accounting profit and money that could, in principle, return to owners or reinvest without new borrowing. Definitions vary slightly by analyst. Pick one, apply it consistently, and note one-time distortions instead of pretending they did not happen.

A research file should show the direction of margins across a cycle, not a single heroic year. If the industry is cyclical, last year's margin may be a peak. If the company is investing heavily, today's margin may be depressed on purpose. Both cases are researchable. Neither is a reason to skip the table. You are trying to answer whether the economic engine still works when the slide deck is less flattering. That answer lives in multi-year statements, not in a single quarter's surprise.

Debt and obligations decide who gets paid first

Leverage is not inherently reckless. It is a claim on future cash that ranks ahead of yours. Read the debt schedule, the covenants if they are described, the lease obligations, and any off-balance commitments the footnotes flag. A company that can service its obligations through a dull stretch of earnings is a different holding than a company that needs the next refinancing to go well. Interest coverage and the mix of fixed versus floating rates are part of that picture, especially when rates have already moved.

Do not stop at the ratio a screener printed. Look at maturity walls: large amounts coming due in a short window can turn a manageable leverage story into a negotiation. Look at who the lenders are only insofar as the filing tells you. Look at whether management has a habit of adding debt to fund buybacks at high prices. The point is not to become a credit analyst. The point is to refuse a common-stock thesis that only works if the creditors stay friendly forever.

Valuation is a chapter, not the cover

After you understand the business, you still have to ask what you are paying. Multiples of earnings, sales, or cash flow are shorthand comparisons, not laws of nature. A discounted-cash-flow sketch is a way to make assumptions explicit, not a machine that prints a true price. The companion article on overvalued and undervalued stocks goes deeper on those tools and on why a low multiple can be a trap. In the research file, valuation belongs after the business description so that you do not reverse-engineer a story to justify a number you already like.

Write down the multiple or the assumption set you are using and the peer set you consider fair, including why those peers belong in the set. If your thesis requires the multiple to expand without an improvement in the business, say so in the file. That is a sentiment bet, and it should be sized like one rather than like a core holding. Research that skips valuation is incomplete. Research that starts with valuation and hunts for a narrative is marketing you are doing to yourself, and it will be especially convincing on days the price already moved your way.

Competition, management, and industry sit outside the spreadsheet

Financial statements describe the past with rules that you can check. Competitive advantage describes why the future might look like a continuation rather than a coincidence. Revisit the advantage paragraph you wrote when you screened the name. Has a rival shipped a substitute? Has a customer built the capability in-house? Has regulation shifted the field? These questions are qualitative and still mandatory. A model that assumes perpetual share gains without naming who loses those shares is a wish, and wishes do not belong in a file you will reread after a decline.

Management quality is easy to overfit to charisma. Prefer evidence you can cite: capital allocation across a cycle, honesty in describing setbacks, insider ownership that aligns without becoming a governance risk, and related-party dealings the proxy statement discloses. Industry structure belongs in the same section. A capable team in a structurally lousy industry can still struggle to produce owner-like outcomes. You are not grading people. You are grading the setup they operate in and the choices they have already made with owners' capital.

• Restate the advantage without marketing language
• Name two rivals and what they could do that would hurt
• Read the proxy for incentives and related-party items
• Describe the industry's pricing power in a sentence

Read the statements, then the ratios, then the peers

The annual report, the quarterly updates, and the footnotes are the primary sources; everything else is a compression or a commentary. Ratios are compressions of those sources. Peers tell you whether a number is unusual. Work in that order even when a dashboard is more entertaining. Starting with a ratio screen is how people debate price-to-earnings while missing a restatement, a contingent liability, or a customer that is walking away. SEC filings exist so that you do not have to take a summary as the whole story, including a summary produced by software.

When you compare companies, compare like with like rather than like with a logo that happens to trade nearby. Different revenue recognition, different capital intensity, and different stages of the cycle will make identical ratios misleading. A peer set of three to six names you can actually read is more useful than a bank of twenty you will not open. Note where the company is better, worse, and merely different. Different is not a defect. Unexamined difference is, because it is where you accidentally pay a premium for a business you did not intend to own.

Risks, mistakes, and a checklist that slows you down

Every research file needs a section titled what would prove this wrong, written before the price tests you. Customer loss, margin collapse, a failed product cycle, a regulatory hit, a refinancing scare, or a key-person departure are typical families. Pick the ones that match the business rather than copying a generic list. Then decide in advance what you would do if they arrived — not a price target, a process: reread, reduce, or exit. Deciding that while you are calm is the entire point of writing it down, because calm is not the default after a gap down.

Common research mistakes are predictable enough that you can put them on the wall. Anchoring on the first number you saw. Ignoring dilution. Treating adjusted earnings as cash. Skipping the footnotes. Confusing a good product you use with a good business at this price. Forgetting overlap with an index fund you already hold. Outsourcing the conclusion to a personality. The checklist below will not make you brilliant. It will make it harder to skip the parts that usually matter, which is a more realistic ambition for a file you will actually keep.

• Thesis in one sentence, including why now
• Revenue mix and customer concentration in your words
• Margin and free-cash-flow trend across several years
• Debt, leases, and near-term maturities
• Valuation method and peer set, written down
• Advantage, rivals, and industry structure
• Disconfirming events and a precommitted response
• Portfolio fit: size, overlap, and what you are funding it with

Put the file next to the portfolio before you decide

A complete research file can still be a poor purchase if it duplicates a giant you already hold through a fund, if it would make one sector dominate the mix, or if it would use cash you need inside a year. Portfolio fit is not a lesser chapter. It is the chapter that turns a good business at a fair price into a good decision for this household. Write the intended weight, the source of funds, and the overlap in the same document as the margins. If those lines are blank, the file is a book report.

This is also where a second set of questions helps. StockLift's analysis tools can surface overlap and structure the pre-purchase checklist. They do not replace the statements and they do not send an order. Use them after the file exists so that you are interrogating a document, not asking a tool to invent one. Then wait a day on anything large enough to matter. Ideas that require immediacy are usually ideas that have not survived contact with the overlap line. Immediacy is not a research conclusion.

AI can accelerate the reading. It cannot own the judgment.

Language models are fast at summarizing a filing, listing risks a company already disclosed, and turning your notes into a tighter outline. They are also confident when they are incomplete. They can miss a footnote, flatten a segment that needed more weight, or treat a marketing claim as a fact because it was repeated in the source. Use them as a second pass after you have opened the statements, not as a substitute for opening the statements. If the tool cannot point you back to the paragraph in the filing, treat the output as a prompt to go look, not as evidence.

StockLift's AI Trade Checker and related Learn tools are built around that slower workflow: thesis, size, overlap, and what would prove you wrong. They do not issue a buy rating and they do not send an order. Use them to keep the file honest, then make the decision at your brokerage. Research that ends in a checklist you can reread is complete enough to act on or to reject. Research that ends in a vibe you cannot source is not research yet.