Building a Research Process for Portfolio Strategy

Start with a question the portfolio can answer

A research request should name a decision boundary, not ask someone to “analyze the market.” In portfolio strategy, ask which exposure, cash constraint, or source of risk needs review and what information could resolve the uncertainty. Whether two holdings overlap is a different research question from whether a valuation input has changed.

Set the scope before opening a chart: relevant holdings, measurement date, comparison set, and observation period. Match the method to the question and keep missing information visible. A dashboard can organize evidence, but its indicators do not define what the portfolio needs to know.

Build an evidence log that preserves context

For each material claim, record the source, retrieval date, metric definition, unit, and what the source directly supports. Separate reported information from estimates and from the researcher’s interpretation. If a source covers broad allocation principles but not a particular instrument, do not use it to validate that instrument’s current spread or expected return.

Mark a model output as an observed figure, a scenario, or a chosen input, since those labels carry different degrees of support.

• Claim, source, and observation date
• Metric definition, denominator, and period
• Assumptions and unresolved evidence gaps

Make the review cycle answer a new question

A review cycle is useful when it revisits material assumptions, changes in the portfolio, or newly available evidence. Its timing can depend on the mandate, account reporting, and the kind of risk under review; there is no single cadence that fits every portfolio.

At each pass, compare like-for-like records and note whether a conclusion changed because the facts changed, the method changed, or a data source was revised. If no new evidence addresses the research question, document that rather than manufacturing a new signal.

Leave a decision record, not a bare verdict

The decision record should summarize the question, evidence considered, assumptions, alternatives, and remaining uncertainty. It should also state what observation would prompt another review, so later readers can distinguish a planned check from hindsight.

A disciplined workflow makes the analysis reproducible without guaranteeing an outcome or dictating an allocation. Retain the evidence log and its limits alongside the conclusion, especially when information could be stale or incomplete. A documented research decision makes a review repeatable, while leaving the conclusion open to new evidence.