What do you actually need to know?

What Do You Actually Need to Know?

An intelligence request framed as �tell me everything about this market� has no natural boundary and rarely improves a decision. Start with a choice, a deadline and a consequence: what must be decided, by when, which actions are available and what would be costly to get wrong. Research has value when it reduces uncertainty separating those actions.

Convert the decision into a few intelligence requirements. Each should specify the unknown, population, geography, horizon and precision. �Will demand grow?� becomes �Which two segments could add enough contribution within 24 months to justify capacity, and what evidence would disprove that case?� The sharper question eliminates data that cannot affect the commitment.

Build the evidence plan before searching. Separate verifiable facts, estimates requiring models and judgments requiring interpretation. Assign the best source to each claim, define corroboration and state the confidence needed. Primary data is not automatically superior; its value depends on coverage, incentives, recency and whether its collection fits the question.

Good tradecraft exposes what is missing. The US intelligence community�s ICD 203 requires analysts to describe source quality, distinguish information from assumptions, explain uncertainty, consider alternatives and address customer implications. The disciplines translate directly to commercial intelligence: an answer without provenance or a competing explanation has not survived scrutiny.

Finally, set a stop rule. Research ends when remaining uncertainty no longer changes the preferred action, the next evidence costs more than its likely decision value or the deadline requires a reversible test. The deliverable is a concise judgment, confidence level, decisive evidence, alternative explanation and the signpost that would trigger reassessment.