Investment process
Keeping a research record that can be revisited
A research record links facts, assumptions, evidence and revisions so that conclusions can be revisited and improved.
01Record for future reconstruction
Preserve the object, question, information date and conclusion. IPO document versions and instrument terms matter. A record should let another reader reconstruct the reasoning, including missing information and unverified claims, rather than merely store files.
02Preserve original provenance
Keep document dates, periods and specific source locations. Distinguish company disclosures, industry context and event reporting. Reposts are not independent evidence. Public references should be relevant and proportionate; analysis adds value beyond assembling links or copied passages.
03Separate facts and assumptions
Plans are not sales, and market growth is not company profit. Preserve intermediate conditions between a technical or industry observation and the investment conclusion. Competing explanations can be recorded with a test for distinguishing them. Precision about uncertainty is not vagueness.
04Keep definitions and calculations
Record units, currencies, periods, scope and numerator/denominator choices. Preserve adjustments and enough workings for recalculation. Changes in disclosure or consolidation can break a series. A neat chart should not conceal incompatible data.
05Specify operating checkpoints
Link each thesis to suitable milestones for customers, construction or economics. Set realistic time frames without making long-term plans untestable. Continue after listing through capital use and operating evidence, not only allocation or first-day performance.
06Keep contrary evidence visible
Make disconfirming information a required part of the record. Apply the same verification standard to favourable and adverse material. Preserve real disagreements and predefine observations that would undermine a key explanation. A useful record supports correction, not self-justification.
07Explain revisions with versions
Log new evidence, calculation corrections and changed judgements separately. Match revision size to evidence quality. Preserve valid components when another assumption fails. A reasoned change is disciplined; unexplained persistence or revision is not.
08Maintain consistency across languages
Chinese and English may differ in expression but must preserve facts and risk boundaries. Corporate registration, regional presence and qualifications do not prove investment returns. Use concrete business language and structured reasoning rather than unsupported claims or jargon.
09Keep a correction trail
If a customer count covers only one product, preserve the old definition, corrected source and affected conclusions. Recheck downstream comparisons rather than replacing the number alone. Document how the process will detect similar errors next time.
10Build a maintained knowledge base
Articles should ask distinct questions while sharing a consistent method. Revisit sources, facts and interpretations as information changes. A reliable record cannot guarantee correct judgements, but makes both successes and errors understandable and future research cumulative.
- Document reference
- Research note · R-09
