Not a versus piece, despite the title: two structurally different tools, each right for different jobs. How accumulation and derivation differ, what each cannot do, and a decision framework — including the rows where the database wins.
Most sourcing disappointments come from asking one tool to do the other's job. Two structurally different objects exist — each with real strengths — and the trade-offs are architectural, not vendor-specific.
A curated index: profiles assembled over years from filings, registrations, news, funding events, and web crawls — wrapped in workflow (search, filters, alerts, contact data, CRM sync).
The entire domain population of a category, read and classified against a specific buyer's thesis at the moment of screening — existence-driven, not source-driven.
The share of the real, operating population a tool can show you.
How much verified information the tool holds per company. The two are in tension in every data product ever built.
Neither architecture is superior in the abstract — each buys something the other structurally cannot reach.
Credit where due. If your targets are $20M-revenue-plus companies with institutional footprints, a good database probably already covers most of them with data no website read can reconstruct.
For companies that generate filings, transactions, and news, profile quality is genuinely strong: firmographics, ownership histories, comparable transactions, executive contacts.
Saved searches, trigger-event alerts, contact enrichment, CRM integrations, team collaboration. A census is an episodic artifact — not a login your associates live in.
The same subscription serves sourcing, comps, market landscaping, and contact finding. A census is thesis-specific by design — no pretense of general-purpose reference.
Databases invest heavily in verified emails and direct dials; web mapping yields whatever contact paths companies publish. For list-to-dialer workflows, that difference matters.
Four limitations are architectural — properties of accumulation itself, not of any vendor's execution.
Small private operators generate no filings or coverage — source-driven inclusion misses them. In fragmented verticals this population is most of the market: our specimen census found 17,300 live US companies in a 25,000-domain sample.
Index taxonomies assign codes by heuristic. A misfiled company is invisible to every filtered search. A fifth or more of confirmed fits in our runs lacked the category's obvious homepage keywords.
A real thesis is textual — "independent, field-service-led, ISO 17025 accredited, no group ownership." Most clauses are reading tasks over site language, not database fields.
Ownership changes at small companies surface as quiet footer edits no feed carries. One in ten keyword-perfect candidates proved group-owned at read time. Census evidence quotes live pages — freshness is the run date.
Committees care about this more each year: what does the tool let you prove? For an IC memo or data room, "the database says so" ages poorly when a spot-check fails.
evidence snippets verified in one machining specimen
documented exclusions — a list that can't show what it rejected can't prove what its inclusions mean
when a census errs, the quote shows why and the fix is one row — not a silent model update
The products price differently because they are different economic objects. Neither number is small; the comparison that matters is cost per covered target and cost per supported decision.
Prosecuting one vertical for years — owning the census plus monitoring typically costs less than a multi-seat license and covers companies the license cannot show.
One thesis on personal economics — the difference is sharpest here. See our separate guide on sequencing that spend.
An export that misses a third of a fragmented market doesn't invoice you for the miss. Estimate it anyway — uncontested conversations close on different terms than auctions.
The choice reduces to a handful of questions about your situation rather than the products:
| Your situation | Better-fitting tool | Why |
|---|---|---|
| Targets are established mid-market+, with filings and news presence | Leading company database | Profiles are deep where sources are rich; census adds little there |
| Thesis lives in a fragmented, long-tail vertical | Full-web census | Index coverage is structurally thin exactly where your targets live |
| Criteria are textual: service model, contract language, credentials, independence | Full-web census | Reading tasks, not filter algebra; evidence quoted per company |
| You need contacts, alerts, and daily self-serve search | Leading company database | Workflow is the accumulated product's core strength |
| You must show an IC or client the complete denominator | Full-web census | Enumerated population with documented exclusions; auditable |
| Many ad-hoc questions across many markets | Leading company database | General-purpose reference beats per-thesis projects |
| One vertical, prosecuted for years | Census + monitoring | Own the map; refresh on deltas; re-run ICPs as the thesis sharpens |
These limits are the boundaries that make the rest of the claims trustworthy. The tool that admits what it cannot do is generally the tool whose other claims survive checking.
No revenue, EBITDA, or valuation estimates — only proxy-based size bands. No deal histories, verified contact hierarchies, or longitudinal executive movements.
Reads the web-visible market — in B2B nearly all of it, but purely offline sole operators and closed-network suppliers should be quantified, not waved away.
A few percent of every screen returns insufficient evidence — sites too thin to classify honestly. These land in a flagged category rather than a guessed one.
Superb depth against your criteria — no pretense of general-purpose reference. Re-running a new ICP over an already-mapped universe is fast and included.
Settle it empirically before committing budget. The test costs one vertical and a few hours of comparison work — ideally a market where you believe your current coverage is strong.
Get the full census of a vertical you know well — a proof project, or a manual census if the market is small.
Lay the census against your database export for equivalent filters and your CRM's accumulated records for the space.
Census-only companies are your uncontacted margin. Export-only companies deserve equal scrutiny — misclassified, acquired, or a genuine census miss.
One email. We send the specimen report the same day — every company scored and ranked with full signal transcripts, plus the exclusions we documented and why.
Request the specimen report