Sources, claims and insights: a practical model for market intelligence
A source is where information came from, a claim is one checkable statement taken from it, and an insight is an interpretation built from several claims. Keeping the three apart is what makes research reusable. Most research documents mix all three in one paragraph.
The three records
| Record | What it is | Example (synthetic) |
|---|---|---|
| Source | A document or page with a type, an address and a retrieval date | Meridian pricing page, read 2026-10-09 |
| Claim | One statement, extracted from a source, that can be true or false | Meridian lists the Team plan at $59 per seat per month |
| Insight | An interpretation that depends on several claims and on your context | Meridian is moving up-market; Fieldnote's mid-size agency segment is exposed to a price gap |
Why the split matters
Each record fails in a different way. A source can go stale, because the page changed. A claim can be contradicted by another source. An insight can be wrong even when every claim is right, because the reasoning was weak. If all three sit in one paragraph you cannot tell which kind of failure you have.
Where evidence relationships go
The link between a source and a claim carries a direction as well as an origin. A source can support a claim, contradict it or add context. Those three relationships, recorded per claim, are what let you say that a claim is supported, contested or still unverified.
A worked example
- Fieldnote's product marketer saves four sources: Meridian's pricing page, its changelog, a review-site page and a community thread.
- She extracts six claims, including “Meridian removed its free tier in September” and “Meridian Team plan costs $59 per seat”.
- The pricing page supports the second claim. The thread supports the first but is a single anonymous post, so the claim stays unverified until the changelog or pricing history confirms it.
- She writes one insight: Meridian is dropping self-serve entry. It lists the two claims, notes that one is unverified, and is marked recent because the sources are weeks old.
- She approves it with that caveat and builds a campaign brief from it.
Common ways the model gets broken
- An insight is written first and claims are fitted to it afterwards.
- A claim contains an opinion, such as “Meridian is overpriced”.
- A source is saved as a bookmark with no retrieval date.
- Two claims are joined with “and”, so evidence for one hides the other.
What to keep for each record
- Source: type, address, retrieval date, the excerpt used.
- Claim: the sentence, its source, its status, its evidence links.
- Insight: the claims it uses, a confidence label, freshness, reviewer and review date.
Who should own each record
Sources are cheap to collect and anyone can do it. Claims need a careful reader, because a badly worded claim misleads everything built on it. Insights need someone who knows the business and will be accountable for the interpretation. Giving the three jobs different owners, even on a two-person team, makes it harder for an interpretation to slip in as a fact.
The same model in Sigblaze
Sigblaze stores all three as separate records on an intelligence graph, with lineage from source to claim to insight to action, and you can trace it in either direction.
This piece contains no external statistics. Fieldnote and Meridian are fictional companies, and the figures in the examples are synthetic.