Build a research note you can challenge

Start with a question, not a favorite ticker

Research becomes easier to evaluate when it begins with a specific question. 'Is this company good?' is too broad. 'Why has cash collection weakened while revenue grows?' creates a testable investigation. The goal is not to gather enough attractive facts to justify an existing preference. It is to understand the business, the claim being purchased and the uncertainty that remains.

First confirm the legal issuer, share class, listing venue and reporting currency. Similar names and multiple listings can lead to mistakes. Distinguish the operating company from a fund, depositary receipt or derivative referencing it. Then define the time horizon of the question: a short-term execution problem and a five-year business forecast require different evidence.

Build an evidence ladder

Original filings and issuer announcements anchor the record. For U.S. reporting companies, EDGAR provides access to filings such as annual, quarterly and material-event reports. Equivalent disclosure systems exist elsewhere. Check publication dates, reporting periods and amendments. A document can be authentic yet stale, and an old figure can become misleading when displayed beside a current share price.

Secondary commentary can identify useful questions, but trace material claims back to their evidence. Keep direct observations separate from calculations and judgments. 'Receivables rose 20%' is a calculation; 'customers are struggling' is a possible interpretation. An alternative could be a major late-period sale on ordinary credit terms. Good notes preserve that distinction rather than compressing it into a confident story.

Write the opposing case

A compact research note can include business model, unit economics, financing, valuation assumptions, catalysts and risks. For every favorable claim, ask what observation would weaken it. If a thesis depends on repeat purchases, customer retention data matters. If it depends on refinancing, debt maturities and funding conditions matter. A risk section should connect to the actual argument rather than repeat generic warnings.

Consider what is already embedded in the price. An attractive business can deliver a poor result when purchased at expectations it cannot meet. Conversely, an inexpensive-looking multiple may reflect a real deterioration. Scenario analysis should vary the important drivers and show the effect on the claim, not merely rename almost identical forecasts 'bear', 'base' and 'bull'.

Make the process reviewable

Date the note and record sources beside the statements they support. Keep a short list of unknowns and planned checks. After new information arrives, preserve the old thesis rather than rewriting history. Reviewing what was believed at the time helps distinguish bad reasoning from a reasonable judgment followed by an unfavorable outcome.

Include implementation constraints even in a paper exercise: spread, likely liquidity, account currency, fees, taxation uncertainty and instrument terms. Research does not oblige a transaction. 'Insufficient information' is a legitimate conclusion. This workflow is an educational method for organizing evidence, not a personalized security recommendation or a promise that more research removes the possibility of substantial loss.

Keep a small evidence register

An evidence register can be more useful than a folder full of unread reports. Give each important claim a source, reporting period, calculation and unresolved limitation. For example: “Operating margin increased from 12% to 15%; calculated from the company's annual statements; includes an acquired division for six months.” The final clause matters because it stops the ratio from quietly becoming a claim about improvement in the original business.

Keep the original number next to any adjustment. If a reported profit includes a property-sale gain, preserve both the reported total and the version excluding that gain, with the arithmetic visible. Someone else should be able to reproduce the result without guessing which rows were removed. This is also useful when the same researcher returns months later and no longer remembers why a particular number looked obvious.

For copied tables, check a few figures against the actual filing rather than assuming extraction was perfect. A lost minus sign, a misplaced column or a unit change from thousands to millions can overwhelm otherwise careful analysis. Record amended filings as new evidence and note which earlier conclusions they affect. An amendment should not disappear into the folder as though the original information had always been correct.

Choose the next question by its effect on the conclusion

Research has an opportunity cost. If a valuation barely changes when an immaterial expense moves, another hour refining that expense may be less useful than resolving a major debt maturity or customer-retention assumption. Rank the unknowns by how much a plausible answer would change the conclusion and by whether better evidence is realistically available. Some uncertainty can be reduced with a filing; some concerns a future competitive response that no document can settle.

Consider a fictional business whose shares appear inexpensive only if a new factory reaches full capacity within twelve months. Reading more commentary about its brand may feel productive, but the decisive questions concern commissioning, customer commitments, unit costs and the cash available during ramp-up. A useful next step targets one of those uncertainties rather than collecting another favorable quotation.

Set a review trigger for each major assumption: a specified operating update, a financing announcement, or the next comparable-period cash-flow statement. Keep outcome and process judgments separate when reviewing later. A price rise does not retroactively validate a weak calculation, and a price fall does not prove that every original assumption was unreasonable. The record should reveal what changed, what was missed and what the evidence still cannot establish.