Competitive intelligence relies entirely on public information. That is not a limitation imposed by ethics alone — it is also that the public record is far richer than most people assume, and the discipline is knowing which parts of it carry signal.
Below are the sources that reliably repay attention, what each is good for, and where each will mislead you.
Product and pricing surfaces
Pricing pages. The highest-density strategic document most companies publish. Tier boundaries reveal who they want as customers, what they consider premium, and where the upsell pressure sits. Changes over time are even better than the current state — archived versions show strategy shifts before anyone announces them.
Documentation. Written for people who already bought, so it is honest about limits in a way marketing never is. Also reveals architecture, integrations, and constraints. Consistently underused.
Changelogs and release notes. Velocity and direction. A changelog that goes quiet for two quarters says something; one suddenly full of enterprise features says something else.
Status pages. Public incident history, including how they communicate under pressure.
Organisational signals
Job postings. The best leading indicator available, and the most consistently overlooked. Roles describe what a company is building six to twelve months before it ships. Volume by function shows where investment is going; a burst of enterprise sales roles precedes an upmarket move; a compliance hire precedes a certification push.
Read the requirements, not just the titles — a listing naming specific technologies tells you what they are building on.
Leadership changes. Departures cluster before pivots and after bad quarters. A new executive from a particular kind of company signals the direction they were hired to pursue.
Employee review sites. Biased toward the disgruntled, and still useful for detecting sustained internal strain, which eventually reaches customers.
Financial and legal record
Company registry filings. In most jurisdictions, registered address, directors, filing history and sometimes abbreviated accounts are public and free. Late filings are a real signal.
Funding records. Amounts matter less than timing and investor type. The gap since the last raise, measured against category norms, tells you whether a company is approaching a decision point.
Patents and trademarks. Slow-moving and occasionally decisive. A trademark filing for an unannounced product name is about as clear a signal as the public record produces.
Customer voice
Review sites. Free-text over ratings, three-star reviews over five-star and one-star, recent over voluminous.
Community threads. Unprompted assessments, in the buyer's own vocabulary. Covered in listening where you don't own the channel.
Their case studies. Marketing, obviously — but the customers named are real, and the pattern across them reveals which segments they actually serve well enough to reference.
What each source is bad at
Every source has a systematic bias, and knowing it is what separates research from collection:
- Pricing pages hide negotiated reality. Published enterprise pricing is often fictional.
- Job postings include aspirational roles that never get filled, and roles kept open to collect resumes.
- Reviews over-represent extremes and lag product changes by months.
- Funding announcements are marketing events, timed for effect, sometimes announcing money raised long before.
- Case studies are the best outcomes, selected.
- Registry filings can be a year stale by the time they publish.
The defence is triangulation. Any single source can mislead; three independent sources pointing the same way rarely do.
The discipline that makes it usable
One habit matters more than any source on this list: record the source with the fact, at the moment of capture.
An unattributed claim found three weeks later is worthless, because you cannot tell whether it came from a filing or from someone's speculation in a forum, and you will not reconstruct which. Every finding should carry where it came from and when it was captured.
The second habit is separating observation from inference. “They posted eight support roles in Q2” is an observation. “They are struggling with support load” is an inference. Both belong in the analysis; conflating them is how a reasonable guess becomes an unchallengeable fact three meetings later.
Where the line is
Competitive intelligence is public-record research. It is not pretexting, not misrepresenting yourself to obtain information, not soliciting confidential material from employees, and not accessing anything requiring circumvention.
The practical test: could you describe your method to the competitor without embarrassment? Everything on this list passes. Anything that does not, is not intelligence work.
Our Competitive Intelligence report draws on these sources with each finding cited and inferences labelled as inferences, so a reader can weigh the evidence rather than take the conclusion on trust.