Lin asks an AI assistant to research a company. Minutes later, forty pages arrive: promising industry, rising revenue, and a target price. At the end, she still cannot say which claim matters most or what evidence would overturn the recommendation to buy.
The problem is the missing connection between evidence and judgment. How does the company earn money? How much growth becomes cash? What expectations does the price already contain? Can the investor accommodate being wrong? Follow one question through that chain before trying to produce a professional report. Later chapters teach the individual business, financial, and valuation methods.
What you will be able to do
- Identify the inputs and outputs needed at each stage of company research.
- Locate missing links between a business explanation, financial evidence, and a price judgment.
- Set conditions for stopping, continuing, and reviewing the work while preserving its versions.
Put a question on the first page
River Mist Software and every number in this example are independent synthetic teaching inputs. Its hypothetical annual report gives current revenue of $2.4 million versus $2 million previously. Year-end receivables rise from $300,000 to $700,000. The largest customer accounts for 45% of current-year revenue. No profit, cash-flow statement, contract terms, or stock valuation is supplied.
Revenue growth is 20%, a result you can recalculate. It raises two questions rather than answering the investment case: why did revenue increase, and did collection keep pace? Higher receivables do not automatically make revenue fictitious. Payment terms, delivery timing, and other circumstances could matter. One balance cannot identify the cause.
Lin narrows her question: “Did this year's growth depend on more generous payment terms, and what would that imply for future cash and the price I would consider paying?” This is a manageable investigation into disclosed business and collection evidence. It is not a forecast of next month's share price.
She limits the project to four hours this week, with no security purchases. The deliverables are a one-page question record, a source index, and a next-action list. If reliable collection evidence is unavailable, she will end this round with the explanation unresolved. Producing a buy recommendation is not a completion condition.
Scope may change, but the reason should remain visible. A later price rally does not let Lin quietly replace “check collections” with “find any growth story” and declare the original task complete. The saved question makes that change detectable.
Give each document a job
Annual filings, subsequent disclosures, and investor-relations materials provide starting points. EDGAR offers access to public filings, including amendment identifiers.[1] Obtaining a filing does not mean a regulator has endorsed the analyst's conclusion.
For each source, record issuer, document name, covered period, public availability, and relevant location. Attach the supporting passage to the claim it addresses. Twenty links at the end of a report do not show which document supports the explanation of collections. The applicable filing type also depends on the issuer.
Separate the supplied facts from missing inputs. Revenue of $2.4 million and $2 million covers two periods. Receivables of $700,000 and $300,000 are balances at two year-ends. The 45% figure describes customer concentration. Missing items include revenue components, payment terms, receivable aging, operating cash flow, renewal arrangements, and price information.
Lin does not read every report section with equal effort. She first seeks revenue-recognition policies and payment terms, then aging or subsequent-collection information that might distinguish explanations. Did deliveries cluster near year-end, or are customers taking longer to pay? If the documents do not answer, the gap stays open. Another general industry article is less likely to settle that particular question.
Media commentary can suggest where to look. It cannot replace the contractual or accounting evidence. Make the next action specific enough that you can recognize when it has been completed: locating the relevant note is different from adding “read more financial reports” to a list.
Keep the business and financial story on the same track
Explain the business in ordinary language: who pays, for what, how often, why they might stay, and what delivery costs. Software may involve subscriptions, one-time licenses, or services. Each can have different renewal, recognition, and collection implications. Do not fill unspecified fields with industry stereotypes to make a complete-looking diagram.
Industry research then asks whether comparable suppliers changed payment terms, whether customer budgets shifted, and whether growth reflects new demand or market-share gains. Those questions extend the original investigation. They do not require an encyclopedia of every macroeconomic topic before returning to collections.
Financial analysis tests whether the explanation connects to the statements. Revenue is a flow over a period; receivables are a balance at a date. Subtracting the change in receivables from a revenue figure does not, by itself, produce operating cash flow. Full statements, notes, consistent definitions, and other cash movements are needed.
The SEC's filing-reading guide identifies business, risks, management discussion, financial statements, and notes as different sources of information.[2] Use that map to identify the missing material. Three supplied figures do not justify an accounting diagnosis or a fraud allegation.
Keep the period labels beside each financial input. A year-end receivable is what remains unpaid on that date, while annual revenue covers activity across the year. A larger ending balance can prompt questions about collection, but it does not identify the timing or cause of every cash movement during the period. That is why the next document request needs both accounting detail and collection evidence, rather than a single subtraction presented as a complete cash-flow result.
The largest customer's 45% share brings the work back to the business. Does that customer have longer payment terms or an approaching renewal? Losing it could affect sales and collections together. Concentration is not an automatic prohibition on investing. It is a reason not to model every revenue dollar as an independent, stable customer relationship.
Business quality still leaves a price question
Suppose later evidence establishes that revenue is real and collections are normal. The stock's attractiveness remains unresolved. A price may already require unusually strong growth. Conversely, a low quoted share price does not establish that a troubled business is cheap relative to its value.
Valuation needs methods and inputs suited to the business and question. Possible approaches examine future cash, comparable assets, or relevant asset conditions. CFA Institute's public overview connects understanding the business, forecasting, choosing a valuation approach, and communicating the conclusion.[3]
Do not select a universal formula for River Mist here. If cash-flow and capital-structure inputs are missing, record them. Multiplying revenue by an arbitrary multiple would conceal those gaps behind a precise target price. Later valuation chapters supply the methods for making the connection responsibly.
You can already write conditional scenarios without pretending to know their probabilities. If renewals proceed and payment terms remain stable, which revenue and collection assumptions could follow? If the principal customer delays payment, which funding requirements change? The scenarios should represent genuinely different conditions, not three numerical decorations around the same optimistic story.
Counterevidence must be capable of changing the explanation. Lin initially suspects looser payment terms. Reliable evidence of unchanged terms and year-end delivery concentration would require her to revise that suspicion. A separate page titled “risks” does not test the central claim if none of its contents could affect it.
A useful conclusion can be a stopping point
Lin can currently write: “Reported revenue rose 20%, year-end receivables increased $400,000, and the largest customer supplied 45% of revenue. Payment terms and cash-flow information remain missing. Sustainable cash conversion and valuation cannot yet be assessed.” That is a bounded interim result, not a failed attempt at a buy recommendation.
A research report should allow its analysis, forecasts, valuation, and recommendation to be checked together, with important assumptions and risks visible.[3] A personal record can be much shorter than an institutional report. It still needs the conditions most capable of changing the judgment.
Even sufficient research does not determine a purchase amount. Household funding needs, existing positions, and risk constraints remain separate inputs to an investment action. “Worth investigating” cannot fill in an allocation, and a valuation estimate is not an order authorization.
At the end of four hours, Lin records “waiting for collection evidence; no action.” She does not extend the report indefinitely because more pages are easy to generate. If public materials cannot reasonably address the question, archiving is available. If a relevant update is expected, she records the event that would reopen the work.
An explicit stopping condition prevents two errors. It avoids calling an unsupported conclusion complete, and it avoids treating every unanswered question as a demand for unlimited effort. The current deliverable can be an evidence gap with a clear consequence for the decision.
Review the changed link, not the whole company introduction
When the next relevant disclosure arrives, compare it with the saved question. Does it alter evidence, an assumption, or a valuation input? Revenue, contract amendments, and major-customer developments may return the investigation to different stages. There is no need to copy the company description into another forty-page report.
If the major customer has left, revisit business and cash scenarios before valuation and portfolio implications. If only the market price changed, first distinguish new operating information from a different relationship between price and your earlier valuation assumptions. The process can return to an earlier step instead of ending forever after one completed pass.
Preserve each version and its information cutoff. River Mist's next-entry point might read: “If the next disclosure provides payment terms, aging, or collection information, check the collection explanation before starting valuation.” This tells a future reader where to resume. “Keep monitoring developments” leaves that choice unresolved.
The record also separates later knowledge from what was available earlier. An updated judgment can be better informed without making the old version dishonest. Overwriting the original would remove the evidence needed to understand why the judgment changed.
Connect the steps yourself
Explanation question: Why does verified 20% revenue growth still fall short of a buy conclusion?
Application question: A report says, “Receivables increased $400,000, so operating cash flow fell $400,000.” Which stage should receive this claim back, and what material is missing? Write your answer before continuing.
For the first question, distinguish the growth mechanism, conversion into cash, price expectations, and the investor's constraints. For the second, return to complete financial statements and notes. Check periods, account definitions, and other cash movements. Revenue and two receivable balances alone establish neither the change in operating cash flow nor fraud.
Change the business, then change the evidence request
A separate fictional manufacturer reports no revenue growth and a large new equipment investment. You are not told when the equipment starts operating or whether future orders exist. Your hypothesis is, “The equipment will substantially improve cash next year.” Identify the first two kinds of evidence you need and a state in which this round can end.
Begin with the equipment's purpose, construction and operating schedule, and with demand, orders, and funding commitments related to it. Capital spending is not automatically an increase in current earnings. Maintenance, working-capital needs, and financing conditions may matter too, but missing evidence is not a reason to invent a complete valuation model now.
If reliable operating and demand evidence cannot be obtained within the four-hour budget, a valid outcome is: “Insufficient support; pause valuation and revisit after project-progress disclosure.” The same research chain applies, but its documents change. Copying River Mist's receivables checklist would not answer the equipment question.
Save a map of the work's current state
Choose River Mist or another fictional business. Create entries for question and scope, sources, business and industry, financial checks, valuation and counterevidence, conclusion and constraints, and monitoring or archive. For each, record available input, current output, the next document needed, and any unmet completion condition. “Not started” is more useful than an invented answer.
Identify one critical gap, one stopping condition, and one event that triggers review. Another person should be able to tell what you would investigate in the next hour and what new evidence would change the conclusion. If the answer is only “study reports” or “watch risk,” narrow it to an identifiable document, field, or event.
Make each step support the next
Turn interest into a bounded question, then let that question direct evidence gathering. Business explanations must survive financial checks; operating judgments must still connect to price and scenarios. Where a link is missing, stop the conclusion at what the evidence supports.
Stopping conditions give the process an exit, and review events give it a return point. The durable output is a record that can explain what changed and why when new information arrives, rather than an opinion that must be rewritten from memory.
References
- Investor.gov: Using EDGAR to Research Investments. Public filings and amendment identifiers.
- SEC: How to Read a 10-K. Uses of business, risk, discussion, statements, and notes; not a complete current filing-rule checklist.
- CFA Institute: Equity Valuation: Applications and Processes. Public overview and summary only; no paid curriculum access.
Sources checked October 30, 2026. AI-assisted draft; a separate AI reviewer has completed an independent full-text review; human financial review and target-reader testing remain pending. River Mist and the manufacturer are original synthetic teaching inputs.