A startup does not need an intelligence department. It needs a small, repeatable system that protects important decisions from stale assumptions. Monitor three to five companies, attach each source to a decision, review only material evidence once a week, and close the loop by recording what changed in the business.

What this guide covers

  • the minimum viable competitive-intelligence system;
  • which decisions and sources deserve scarce attention;
  • a 45-minute weekly review;
  • a 30-day rollout and operating ledger.

Begin with recurring decisions

Choose no more than four decision classes:

  • packaging and pricing;
  • product boundary and integrations;
  • positioning and sales comparisons;
  • enterprise, reliability, or risk readiness.

For each class, write an owner and an action threshold. “Track Cursor” is not a decision rule. “If a direct coding-agent competitor changes included usage or team controls, the founder and product owner review our plan boundary within two business days” is.

Official sources should match the field. Cursor pricing supports current public offer fields; Cursor status supports disclosed service incidents. Neither source establishes adoption, motive, or overall product quality.

Keep the peer set painfully small

Select three to five direct companies seen in actual deals, customer interviews, roadmap debates, or budget substitution. Add one adjacent company only if its movement could alter the same buyer decision.

The EXVIV company directory and market map help discover the broader landscape. Your operating set should be smaller. Breadth is useful for exploration; relevance is necessary for action.

Review membership quarterly. Remove a company when the overlap disappears. A famous company does not deserve permanent attention merely because the team knows its name.

Build one source card per company

# Company source card

Why this company matters:
Affected buyer/job:
Decision owner:

| Source | Fields watched | Cadence | Action threshold |
|---|---|---|---|
| Pricing | plan, unit, allowance, entitlement | weekly | material package movement |
| Changelog | launch, availability, deprecation | weekly | buyer-job overlap changes |
| Homepage | audience, promise, proof | monthly | persistent positioning movement |
| Status | component, severity, duration | event-led | recurring relevant pattern |

Last peer-set review:

Do not add jobs, trust, social, and every documentation page automatically. Every source consumes review capacity.

Use a 45-minute weekly review

Ten minutes: source health

Check failures, redirects, blocked pages, and incomparable captures. Source faults are not competitive events.

Fifteen minutes: evidence review

Inspect exact changed fields and before/after context. Reject navigation, timestamp, experiment, and formatting noise.

Ten minutes: implication

For each material item, separate observation, plausible implication, alternative explanation, and unknown.

Ten minutes: routing

Assign one owner, one review, and one due date—or record “no action.” The system is incomplete until the item reaches a terminal decision.

Preserve an action ledger

Evidence IDObserved factDecision reviewedOwnerOutcomeRevisit
no action / updated / tested / researched

This ledger prevents the same movement from being rediscovered and allows the team to learn which source types create useful decisions.

Use AI as an assistant, not an evidence source

AI can normalize fields, summarize a diff, group related movements, and draft a brief. It cannot supply a missing before state, prove a vendor outcome, or know private intent. Every material claim should point to a preserved primary source and review decision.

NIST's AI Risk Management Framework emphasizes governed, contextual risk work. Apply the principle narrowly: define responsibility, retain evidence, evaluate performance, and improve the system. Google's people-first content guidance provides a parallel publishing test—original analysis and verifiable sourcing matter more than content volume.

Thirty-day rollout

Week 1: choose decisions, owners, and peer set. Build source cards and establish baselines.

Week 2: run the first review without automating routing. Learn which differences are noise.

Week 3: add structured fields and thresholds for repeated movements. Create the action ledger.

Week 4: review outcomes. Remove sources that produced no relevant evidence, tune thresholds, and document the next quarterly peer-set review.

Use Evidence Compare to inspect public changes and the EXVIV competitor-monitoring guide for the deeper operating method. The practical monitoring article expands the source-to-decision workflow.

Sources and further reading

Method note

This operating system relies on public evidence and internal decision ownership. It does not justify collection behind access controls or conclusions about private intent, adoption, revenue, or product quality.