Foundations
A competitive intelligence process your team can actually run
The classic intelligence cycle still works. The failure usually happens between collection and action, where links pile up and nobody owns the conclusion.
1. Direction: write an intelligence question
Replace broad requests such as “watch our competitors” with a question that has a deadline and an audience. For example: “What changed in enterprise packaging this month, and does sales need a new objection response?”
A good question defines the market boundary and the decision. It also tells the analyst when to stop. Without that boundary, research expands until the answer arrives too late.
2. Collection: preserve the public evidence
Collect from first-party sources before commentary: product pages, pricing, documentation, official accounts, leadership profiles and regulatory records where relevant. Independent reporting and customer discussion add context, but should not silently replace the original claim.
Keep the source URL, visible publication time, collection time and a stable excerpt or snapshot. If a source does not expose a reliable date, record that uncertainty instead of inventing precision.
3. Processing and verification: make unlike sources comparable
Normalize company names, people, dates and event categories. Deduplicate syndicated announcements and repeated social posts. Identity claims—especially founder and executive profiles—need corroboration from the company site or another independent public source.
Verification is not the same as certainty. Use confidence labels that explain what is known, what is inferred and what would change the conclusion.
4. Analysis and dissemination: make the implication explicit
A useful finding states what changed, why it matters to this company, the evidence for the claim and a recommended next step. It should also state when no action is warranted.
Deliver intelligence inside an existing rhythm: a morning brief, launch review, pipeline meeting or product planning session. Then capture feedback. If the audience repeatedly ignores a category, narrow it or remove it.
Working checklist
Use this in the next review.
- Turn the request into a dated intelligence question.
- Prioritize first-party evidence and retain source details.
- Normalize entities and remove duplicated observations.
- Label fact, inference and confidence separately.
- Name the implication, owner and next review date.
- Use audience feedback to change the next collection cycle.
Continue the workflow