How to Analyze a Stock: Business, Financials, Valuation, Risks

To analyze a stock, answer four questions in order: is this a good business, do the financial statements confirm the story, is the price reasonable against the company's own history, and what could realistically go wrong. Work through them in that sequence and you have a complete, repeatable framework. Skip one, and you are not analyzing, you are hoping. This guide walks through each question, shows exactly which numbers to pull and where to find them, and explains how 20 years of price history fits alongside the fundamentals.

Step 1: Understand the business before the numbers

Start by writing one sentence a twelve-year-old could follow: who pays this company, for what, and why they keep paying. For Apple that sentence is easy. For a specialty insurance holding company with three subsidiaries, it might take an hour, and that hour is the analysis. If you cannot write the sentence, you cannot evaluate anything downstream, because you will not know which numbers matter.

Then ask what protects the business. Recurring revenue, switching costs, network effects, brand, regulatory license, or cost advantage: at least one should be identifiable and defensible. A company with none of these can still grow, but its margins are on loan from competitors who have not arrived yet.

Step 2: Check the revenue trend and profit margins

Revenue is the least manipulable line on the income statement, so start there. Pull five to ten years of annual revenue from the company's 10-K filings and ask three things: is it growing, is the growth rate steady or decaying, and did it hold up in the bad years. A company that grew revenue 15% a year from 2015 to 2021 and 3% a year since is telling you something the narrative may not.

Margins tell you whether growth is worth having. Gross margin shows pricing power: software companies commonly run 70% to 80%, grocery retailers 20% to 25%, and the number matters less than its direction. A gross margin that slid from 62% to 55% over four years means the company is buying its growth with price cuts. Operating margin shows discipline: revenue up 40% while operating margin fell from 18% to 9% is a company spending two dollars to look like it earned one.

Step 3: Read the balance sheet for survival

The income statement tells you how the company is doing; the balance sheet tells you whether it can survive being wrong. Three checks cover most of it:

• Debt load. Total debt against EBITDA is the standard yardstick. Under 2x is comfortable for most industries, over 4x deserves an explanation, and the explanation should not be "rates were low when we borrowed."
• Interest coverage. Operating income divided by interest expense. Below 3x, a mediocre year starts threatening the dividend; below 1.5x, it threatens the company.
• Cash conversion. Compare net income to free cash flow over several years. A company reporting $500 million of profit while generating $150 million of cash is booking earnings it has not collected.

Step 4: Value the stock against its own history

Comparing a stock's price-to-earnings ratio to the whole market mostly measures what industry it is in. Comparing it to the company's own 10-year range measures sentiment about this specific business. A quality company trading at 18x earnings when its decade median is 24x is being offered at a discount to its own reputation; the same company at 35x is priced for a future better than its past. Neither is a verdict, but both are information. Do the same with price-to-sales for unprofitable companies and enterprise-value-to-EBITDA for debt-heavy ones.

Valuation is also where price history earns its place next to the fundamentals. A stock's 20-year record, adjusted for splits and dividends, shows how the market has treated this business through at least two full cycles: how deep its drawdowns ran, how long recovery took, and whether today's price sits near the top or bottom of its own valuation range. AgentTrading's historical stock data runs 20+ years of split- and dividend-adjusted daily prices under every analysis, so the history you are reading is the history that actually happened, dividends included.

Step 5: List the risks before you decide

Every stock analysis should end with a written list of what could go wrong, because the risks you name in advance are the ones you can size for. Look for four kinds: concentration (one customer over 10% of revenue, one product over half of profit), cyclicality (did earnings fall more than 30% in 2009 or 2020), balance sheet stress (refinancing due in the next two years), and valuation stretch (priced above its own historical range, so good news is already spent). If your analysis produced zero risks, the analysis is wrong, not the company. This is why AgentTrading's AI stock analysis is required to surface at least one risk flag on every summary: an analysis with no risks listed is an advertisement.

Stock analysis checklist: what to check and where

Where price history fits: fundamentals plus 20 years of record

Fundamentals tell you what the business is; price history tells you what owning it felt like. The two together are stronger than either alone. A dividend blue-chip basket, rebalanced quarterly, looks placid in a filing; run it against 20+ years of adjusted daily data and you see the trade-off drawn honestly: in our historical illustration that basket HELD UP on drawdown and trailed the broad market on growth. Both halves of that sentence matter, and only the price record shows them. If your thesis about the fundamentals implies a testable rule, run it: the fundamental analysis tool turns filings into plain-English summaries, and the same bench backtests the rule with a 0.1% cost per trade assumed by default and every assumption printed on the result.

Past performance does not guarantee future results. For educational and informational purposes only. Not financial advice. Consult a licensed advisor.

Common mistakes when analyzing a stock

• Starting with the price chart. The chart tells you what other people already decided. Read the business first, then let the price record confirm or challenge it.
• One year of data. Any company looks like a trend if you only check one year. Five years is a minimum; ten covers a full cycle.
• Comparing multiples across industries. A railroad at 15x earnings and a software company at 30x may be identically priced relative to their own histories.
• Skipping the risk list. Naming risks after the position moves against you is journaling, not analysis.

Analyze the next stock in minutes, not evenings

The manual version of this framework is a screener, ten filings, and a spreadsheet, roughly 4 to 6 hours per idea. The framework does not change when a tool runs it; the hours do. If you want the same four questions answered with every claim checkable and at least one risk flagged, paste a ticker into the AI stock analysis bench and read what the record says. What you conclude from it is, as always, your call. For a deeper look at what happens after the analysis, when a thesis becomes a testable rule, see how to backtest a trading strategy in 6 steps.

How do you analyze the financial data of screened stocks?

Work in two passes. The screener's job is only to shrink the list, so treat its output as candidates, not conclusions, and never buy from a screen directly. On the second pass, pull three to five years of financials for each survivor and check four things in order: whether revenue growth is real and durable, whether margins are stable or drifting, whether cash flow supports reported earnings, and whether the balance sheet can survive a bad year.

The reason for the two-pass structure is that screeners select on whatever is measurable, which is rarely what matters. A filter for low price-to-earnings will happily return companies whose earnings are about to fall, and a filter for high growth will return companies buying that growth with debt. The financial statements answer the question the filter could not ask. If your candidate list is long, sort it by which names you can most easily disprove and start there, because eliminating quickly is worth more than confirming slowly.

Can you rely on a stock score or summary grade?

Use scores to sort a list, never to make the decision. A composite grade compresses dozens of factors into one number, which is useful for ranking and useless for understanding, since you cannot see which factor is driving it or whether that factor matters for your holding period. A score that says 8 out of 10 gives you no way to check whether it is wrong.

The practical alternative is to keep the score as an input and then verify the claims underneath it yourself. That is the design principle behind the AI stock analysis bench: every line of the summary is a checkable statement drawn from filings and market history rather than an unexplained rating, and at least one risk is always flagged. The honest limits of rating systems are covered in do AI stock pickers work, and once you have a thesis worth testing, how long you should backtest a trading strategy covers how much evidence a rule needs before it means anything. If you are still deciding which kind of software does which part of this job, the category is mapped honestly on stock analysis tools, and a full thesis carried from evidence to verdict is what AI stock research describes.

Analyzing stocks using the Equity Summary Score

Fidelity's Equity Summary Score is a consolidated analyst rating, not a valuation or a signal. It is produced by the LSEG StarMine model, which takes the published ratings from the independent research providers on Fidelity.com, normalizes them so that scarce ratings count for more than plentiful ones, then weights each firm by its own rating accuracy over the previous 24 months in that sector. The output is a single number on a 0.1 to 10.0 scale, updated daily after the close, and it only exists for stocks that at least four firms cover.

What that construction means in practice is worth being clear about. The score summarizes what the sell side currently thinks, adjusted for who has been right lately. It says nothing about what you paid, nothing about your holding period, and nothing about the balance sheet risks in step 3 above. It also has a coverage gap by design: a small or newly listed company with three covering analysts has no score at all, which is exactly the type of name where an outside opinion would be most useful.

The sensible way to use it is as a screening and disagreement tool. Sort a watchlist by the score to see where consensus sits, then spend your time on the names where your own reading of the numbers disagrees with the rating, because that is where the analysis has something to add. Treat a bullish score as a prompt to check why, not as a conclusion, and remember that ratings are opinions with a track record rather than facts. If you want to test a rule built on the score, or on any other ranking, the same evidence standard applies as for any other idea: see stock backtesting for how to run that on one ticker, and is backtesting accurate for how much weight the result deserves.