Competitive intelligence workflow
How to Build a Source-Backed Competitive Intelligence Knowledge Base
Keep competitor research current with a bounded source register, dated evidence, explicit uncertainty, and reviewable knowledge-base Pages.
About this guide
Workflows
Product marketing managers, founders, and product, strategy, or market researchers maintaining competitor evidence
9 min read
01
Start with one decision and a bounded set of competitors instead of an unreviewed market archive.
02
Register each source with its date, revision, scope, and next review so stale claims stay visible.
03
Separate observations, inferences, assumptions, and unknowns before a battlecard or product decision uses them.
Worked example
Build a competitive-intelligence knowledge base
A fictional three-file exercise: compare HarborDesk capability claims with a dated plan, then write the cautious version a buyer can review.
Try this task
Using only the three source files, classify each capability claim, compare it with the dated plan, and write wording that does not imply current availability. Preserve the human review decision and cite each conclusion with [[filename]].
Expected reasoning
Treat SAML, audit-log export, and the rate limit as vendor claims without current verification. The dated plan gives SAML a Q4 2026 target, leaves audit-log export uncommitted, and does not specify the rate limit; a human owner must approve any comparison wording.
Why keep this connected
One fictional competitor can be checked in a few files. A source-linked intelligence wiki helps when dated captures, roadmap changes, and confidence labels must stay connected without turning plans into product promises.
Review boundary
Check each claim against its date and evidence type: the comparison page is a claim, while the plan is not a release. Leave audit export and the rate limit unverified, with wording decided by the comparison owner.
Inspect the example in Wenlan
Wenlan v0.18.3 interface displaying data read back from an isolated test run. The sources are fictional and the reference answer was written for this exercise. No automatic AI generation or approval is shown.
After the cited source changed, Wenlan marked the page as out of date and kept its original text. The ‘updating…’ label indicates a pending rebuild here; it does not show a completed correction. Review the changed source before rebuilding and accepting a new answer.
Source files
Open a file to read its complete authored Markdown.
The source files are the same authored English dataset in all three locale views.
01competitor-claim.mdHarborDesk comparison-page capture
# HarborDesk comparison-page capture Capture date: 2026-09-01. This is a fictional marketing-page excerpt created for teaching. HarborDesk's comparison page says the product offers SAML single sign-on, exportable audit logs, and an API limit of 1,000 requests per minute. The page does not show a plan name, release version, test result, or implementation evidence. Treat each sentence as a vendor claim, not an independently verified capability.
02dated-product-plan.mdHarborDesk product plan — dated internal note
# HarborDesk product plan — dated internal note Note date: 2026-08-20. This fictional note says SAML single sign-on is planned for Q4 2026. Exportable audit logs are discussed but have no committed delivery date. The API rate limit is not specified. A plan is not a release record. No source here verifies that any listed capability is available to a customer on the capture date.
03comparison-review-notes.mdCapability comparison review notes
# Capability comparison review notes The comparison-page capture is the only direct statement of the three capabilities, and it is undated as a product release. The dated plan constrains what can be said about timing but does not confirm delivery. A reviewer may label SAML as a vendor claim with a dated plan reference. Audit-log export and the 1,000-requests-per-minute limit remain unverified. A human owner must decide whether the evidence is strong enough for a comparison table; do not turn the plan into a promise.
Reference answer
Reference answer: separate capability claims from delivery evidence
The comparison page makes three fictional vendor claims, while the dated plan only places SAML in a Q4 2026 plan and leaves audit-log export and the rate limit uncommitted or unspecified. None is a verified current capability; a human reviewer must decide how to present the comparison.
HarborDesk — comparison reference
Evidence status
The 2026-09-01 comparison page claims SAML single sign-on, exportable audit logs, and a 1,000-requests-per-minute API limit. It supplies no plan, release, or test evidence, so these remain vendor claims rather than verified current capabilities. competitor-claim.md
The 2026-08-20 plan places SAML in Q4 2026, discusses audit-log export without a committed date, and does not specify the rate limit. A plan is not a release record. dated-product-plan.md
Warranted wording
A comparison can label SAML as a vendor claim with a dated Q4 2026 plan reference. Audit-log export and the rate limit should be marked unverified. A human owner must decide whether that evidence belongs in a published comparison; no capability should be presented as a current fact. comparison-review-notes.md competitor-claim.md
Changed source: dated-product-plan.md
Expected update after this change
The reference's Q4 2026 SAML plan becomes stale and must change to the revised Q1 2027 target. Exportable audit logs are now explicitly outside the current committed plan, while the rate limit remains unspecified. The marketing claims are still unverified, and a human reviewer still decides whether to publish a comparison.
# HarborDesk product plan — dated internal note, revision 2 Note date: 2026-09-04. This fictional note says SAML single sign-on moved to a Q1 2027 target. Exportable audit logs are explicitly outside the current committed plan. The API rate limit is still not specified. A plan is not a release record. No source here verifies that any listed capability is available to a customer on the capture date.
01
Quick answer
Build one competitor dossier for one decision, review question, or product area. Register user-collected and permitted product docs, manuals, release notes, and research notes with source, date, revision, scope, and review owner. Then maintain a source-backed Page that separates observations from inferences, preserves contradictions, and marks stale claims before anyone uses the dossier.
Wenlan can connect supported Markdown, text, text-extractable PDFs, folders, and read-only Obsidian sources to source-backed Pages, citations, revisions, stale state, lint, and human review. It does not crawl competitor sites, scrape reviews, monitor live pricing, discover competitors, send alerts, score vendors, make recommendations, or decide legal or commercial strategy.
02
When this problem appears
A competitor spreadsheet or battlecard goes stale when product pages, documentation, releases, and positioning change but nobody can tell which claim came from which revision. A polished AI summary can hide that gap. A useful competitive intelligence knowledge base keeps a small, inspectable evidence chain so a researcher can tell what was observed, what was inferred, what is unknown, and what needs another check.
03
Build one competitor dossier that can be reviewed
Use one decision and a small source boundary first. The workflow remains useful without Wenlan because the source register and claim classifications are the durable practice.
- Name the decision, review question, competitors, date range, source types, and excluded material. Do not ask an assistant to discover an unrestricted market corpus.
- Collect only permitted, user-provided sources such as product documentation, user manuals, release notes, dated research notes, or a read-only Obsidian vault. Keep credentials, private customer data, and restricted material outside the boundary.
- Record each source URL or file, publication or access date, version or revision, competitor, scope, authority, allowed use, and next review date.
- Classify every important statement as observation, inference, assumption, contradiction, or unknown. Preserve the exact passage and do not turn an inference into a competitor fact.
- Compare one product question or module at a time across the same source class. Record missing evidence instead of filling a row with generic market language.
- When a release or source changes, resync the affected source and mark dependent claims stale or contradicted until a reviewer checks the new revision.
- Before using a battlecard, positioning update, or product decision, open the source, check its scope and date, and leave the unresolved questions visible.
A bounded competitor-evidence workflow
wenlan status
wenlan sources add ~/Research/competitors
# In a Wenlan plugin client:
/distill <competitor research question>
/pages <competitor topic>
/lint
/curate04
What to check next
This is not automated market intelligence. Wenlan does not crawl or scrape competitor sites, monitor live websites or pricing, ingest reviews, discover competitors, run alerts, score products, generate recommendations, or make legal or commercial decisions. It only organizes supported sources you are allowed to use; verify public facts, permissions, dates, and strategic conclusions with the responsible human reviewer.
Make one competitor decision traceable
Choose a bounded question, connect permitted sources, and review every consequential claim before it reaches a battlecard or product decision.
FAQ