Where growth intelligence data comes from — and what each source can't tell you
Most tools show you a number. Far fewer tell you which source produced it, how current it is, or what it structurally cannot know. This page is the reference for both halves of that: the public sources behind company research, and the limits that come with each one.
Why provenance matters more than the number
A company score with no source trail is an opinion with a decimal point. The moment someone asks "where did that come from?" — a prospect, a manager, a security reviewer — an unattributed figure becomes worthless, because you cannot defend it and you cannot check it.
That is why every figure in LeadScout carries its origin, its method, and its confidence, and why a missing figure is shown as missing rather than filled in with an estimate. The sources below are the ones the product actually queries. Where a category depends on you connecting your own account, it says so.
Company identity and registration
The hardest problem in company research is not gathering facts, it is being certain which company they belong to. Two firms share a name; a typo mints a second record; an abbreviation resolves to a different entity. Identity is confirmed first, once, and every engine reuses it.
| Source | What it can establish | Where it goes silent |
|---|---|---|
| National company registries e.g. India's MCA/RoC open data, UK Companies House, SEC filings | Legal name, registration number, incorporation date, registered address, company class, and — for filers — audited financials. | Coverage and freshness vary by jurisdiction; registry names lag trading names badly, and small private firms may file almost nothing. |
| Wikidata and Wikipedia | Official website, industry, headquarters, parent and subsidiary relationships, founding date. | Only covers companies notable enough to have an entry — which is a small minority of any prospect list. |
| RDAP (domain registration) | Domain creation and expiry dates, registrar, and status — a useful proxy for how long an operation has actually existed. | Registrant identity is redacted for privacy in most jurisdictions, so it identifies the domain, not the people. |
| Company websites | Self-described positioning, products, locations, leadership, and published contact routes. | It is marketing copy. Treat it as what a company claims about itself, not as verified fact. |
Web and technical signals
How a company builds and runs its web presence is observable without asking permission, and it is often more current than anything it publishes about itself.
| Source | What it can establish | Where it goes silent |
|---|---|---|
| DNS records over DoH | Which vendors are in the stack — mail provider, verification records for SaaS tools, CDN and hosting. A strong, current indicator of what a company has actually bought. | Shows a vendor is configured, not that it is used, paid for, or renewed. Absence proves nothing. |
| PageSpeed Insights / Core Web Vitals | Measured load performance and field data for a public URL. | Field data needs enough real traffic to exist; low-traffic sites return lab data only, or nothing. |
| The page itself crawled markup, robots.txt, sitemaps | Technical SEO state: titles, headings, structured data, canonical and index directives, crawlable depth. | Only what is publicly served. Anything behind a login or a paywall is out of scope by design. |
| Web archives | When a site, product page, or pricing page changed — useful for dating a strategy shift. | Archive coverage is uneven and can miss the exact change you care about. |
Hiring and buying-intent signals
Intent is the one category where public data genuinely beats a purchased database, because job postings are published, dated, and specific.
| Source | What it can establish | Where it goes silent |
|---|---|---|
| Public applicant-tracking boards Greenhouse, Lever, Workable, Ashby | Open roles with dates and descriptions — team growth, new functions, and named tools a role requires. A posting asking for a specific platform is direct evidence of that platform. | Only companies using these boards publicly are visible. A company hiring through an agency or its own careers page may show nothing. |
| Vendor-stack changes over time | A new record appearing suggests a recent purchase or migration — a timing signal, not just a state. | Requires repeat observation; a single lookup gives state, not change. |
Contact discovery — and the line we draw
This is where most tools quietly stop being honest, so it is worth stating precisely. Contact discovery here is limited to business contact information a company publishes about itself: a team page, a contact page, a press release, an author byline, a public code repository's commit metadata.
Two things follow from that, and neither is negotiable:
- Nothing is guessed and presented as found. Where an address is inferred from a pattern rather than observed, it is labelled as a pattern with lower confidence — never shown as verified.
- Absence is reported as absence. If a company publishes no contact route, the result is "Not available", not a plausible-looking address.
You remain responsible for how you contact people, and for whatever consent and marketing rules apply where you and they are. A source label is not permission to send.
Your own connected accounts
Public sources plateau. Beyond that point the useful data is first-party, and it is yours, not ours to find: Search Console, Analytics, Ads, and commerce platforms. These are optional, they only ever read what you authorize, and the product works without them — it just shows fewer, better-labelled things.
How confidence is assigned
Confidence is not a vibe. It reflects three things about the evidence underneath a figure:
- Directness — was it observed at the source, derived from something observed, or inferred from a pattern?
- Corroboration — does more than one independent source agree?
- Freshness — how long ago was it actually measured, as opposed to when it was last displayed?
A single inferred signal with no corroboration and an old timestamp gets a low score and says why. That is more useful than a high number you cannot interrogate.
How to check any figure yourself
The test of an evidence-first claim is whether you can go around it. In the workspace: open the figure, read the source and method it names, follow the link to the origin, and check the measurement date. If a figure cannot survive that, it should not be driving a decision — and if you find one in LeadScout that can't, it is a bug worth reporting.
Run it on a company you already know
The fastest way to judge any of this is to point it at a company whose facts you can already verify, and see what it gets right, what it labels uncertain, and what it admits it cannot find.
Try a live search Open the workspace