Research process · 4 min read
Build scenarios you can actually check
What evidence would make you change your mind?
Educational guide · Updated · Prepared with AI assistance
01
Write the observation before the interpretation
Keep three separate lines: what was reported, what you calculated from it, and what you think it could mean. Attach the reporting period and source to the first line. Attach units and a reproducible formula to the second.
For an earnings-led thesis, the issuer's filing and management discussion are starting points. A presentation headline should lead you back to the underlying reporting, rather than stand in for it.
02
Give every condition a definition
Instead of saying margins hold up, specify the company's consolidated GAAP operating margin, the period, and the comparison basis. Instead of cash coverage improves, specify which cash measure covers which spending over what window.
Do not turn an illustrative threshold into a forecast. If you choose a threshold for a research exercise, say that you chose it and explain why crossing it would change the question you are asking.
03
Allow an unclassified outcome
Bull, base, and bear cases do not automatically cover every combination of evidence. Growth can exceed a threshold while cash generation misses another. Label that mixed outcome rather than forcing it into a favorable category.
A review date is separate from a catalyst date. If new information has not arrived by your review date, the honest update may be that the uncertainty remains unresolved.
Hypothetical worked example
A mixed result
In a hypothetical exercise, the constructive case requires revenue growth above 10% and operating margin above 15%, both in the same quarter. Growth of 12% with a 13% margin meets only one condition.
Call it mixed evidence. Investigate the margin change and revise the thesis only when the source supports the explanation; do not assign an invented success probability.
Your next review
- Separate facts, calculations, and interpretation.
- Define the metric, period, source, and threshold.
- State what invalidates the thesis.
- Keep a mixed/unclassified outcome and set a review date.
Source and boundaries
The linked source explains the underlying concepts; the research workflow and hypothetical example are this site’s educational illustration. No live data, expected returns, or personalized recommendation is provided.