Mining Public and Enrichment Data to Build a Company List That Actually Matches Your ICP
Most B2B teams still build lists the same way: buy a database of 50,000 companies filtered by industry and headcount, then wonder why the reply rate sits under 1%. Data mining for cold outreach means the opposite: starting from a narrow, specific set of criteria and pulling only the companies that match, from sources you can verify yourself. It takes longer to build 500 accounts this way than to buy 50,000, but the 500 actually convert.
- Bought databases optimize for coverage, not fit — for cold outreach, a 500-company list built to match your ICP outperforms a 50,000-company list that doesn't.
- Public filings, job boards, and technology footprints are free, current, and specific — they beat static SIC-code databases for finding companies that are in-market right now.
- Enrichment APIs should confirm and fill in a list you've already narrowed with criteria, not generate the list from scratch.
- Verify firmographic and contact data at the point of use, not once at import — companies change size, ownership, and domains constantly.
- Every public data source used for B2B contact data needs a documented lawful basis, captured at the point of collection, not retrofitted later.
Why a Bought Database Is the Wrong Starting Point
Vendor databases are built for breadth: they aggregate everything they can license or scrape, tag it with an industry code and an employee-count bracket, and sell access by the record. That works for reach; it doesn't work for address-based cold outreach, where you're writing to a named person at a specific legal entity based on something true about their business right now. A generic list can't tell you which of 3,000 'software companies, 50-200 employees' just raised a funding round, opened a second office, or posted five open sales-ops roles this month.
The economics also work against you. If a purchased list is a year old, and most are by the time resale licensing catches up, a meaningful share of the domains have changed, the decision-makers have moved on, and the companies that were a genuine fit have since been acquired or shut down. You end up paying for volume and then spending as much time cleaning the list as you'd have spent building a smaller one from scratch.
Data mining, in this context, just means treating list-building as a research task with defined criteria rather than a purchase. You decide what a good fit looks like in specific, checkable terms, then go find companies that match, using sources where you can see the underlying fact for yourself.
Where the Data Actually Lives
The useful sources split into three tiers: public records that are free and authoritative, aggregator platforms that are cheaper than buying finished lists but still need verification, and behavioral signals that tell you who's in-market today.
The order matters. Start from a public or behavioral signal that tells you a company is a plausible fit right now, then use enrichment tools to confirm and fill in the missing fields. Using enrichment tools as the primary discovery mechanism just reproduces the vendor-database problem at a smaller scale.
- National business registries (Companies House in the UK, SEC EDGAR and state Secretary of State filings in the US, similar registries in most jurisdictions) — legal entity name, registration date, filing status, sometimes revenue bands.
- Job boards and company career pages — a spike in open roles for a specific function is a reliable proxy for a team about to scale, which is exactly the trigger point for relevant outreach.
- Technology footprint scanners that show what a company runs on its own website — useful when your pitch is tied to a specific stack or a competitor's tool.
- LinkedIn company pages and Sales Navigator — headcount trend, recent hires in the department you're targeting, and organic posts that hint at initiatives.
- Industry associations, conference exhibitor and speaker lists, and trade press — smaller volume, but very high fit, since attendance already filters for relevance.
- Enrichment APIs layered on top (Clearbit-style tools, Crunchbase, national tax-ID lookups) to fill in firmographic detail — revenue estimate, employee count, funding stage — once you already have a name and domain.
Turning Criteria Into a Repeatable Pull
A mined list needs explicit inclusion criteria before you touch a single source, or the process turns into an unstructured research binge that never converges. Write the criteria down as a short checklist: industry, employee-count range, geography, a technology or hiring signal, and a disqualifier such as already a customer or already on a stop list.
Work source by source against that checklist rather than trying to merge everything at once. Pull 100-200 candidate companies from one source, apply the criteria, keep what clears the bar, then move to the next source and de-duplicate against what you already have by domain, not by company name — name matching produces false duplicates and false negatives once subsidiaries and rebrands enter the picture.
For most cold B2B campaigns, a realistic target is 300-800 tightly-matched companies per sourcing pass. That's small compared to a bought database, but it's the range where a rep or an SDR pod can maintain personalization quality across the whole list without the process collapsing into templated batch sends.
A concrete pull: target is mid-market logistics software vendors, 50-250 employees, US and Canada, hiring for supply-chain roles in the last 60 days. Source 1 (job search filtered by function and posting date) returns 340 companies. Source 2 (a technology scanner filtered for a specific TMS integration) cross-references down to 210 that clear both filters. Enrichment fills in employee count and confirms 165 are still within the size band. That 165-company list is the campaign's target set — small, current, and each row checkable back to a real signal.
Verifying Before the List Touches a Sequence
A mined list is only as good as its freshest field. Verify at three points: the domain resolves and belongs to the company you think it does, the company is still operating at the size band you filtered for, and the contact data attached to it, if you're enriching down to a named person, passes an email-verification check rather than a guessed pattern.
Treat data mining and email discovery as separate steps. Finding the right company is a firmographic problem; finding the right person's working email at that company is a contact-verification problem, and conflating them is how lists end up with a 15-20% bounce rate that then damages sender reputation for every subsequent campaign from that domain.
Re-verify anything older than 60-90 days before reusing it. Company data decays faster than most teams assume — job changes, acquisitions, and website migrations happen continuously, and a list that was accurate in Q1 can have a double-digit error rate by Q3.
Mistakes That Turn a Mined List Back Into a Bought One
The whole point of mining data source by source against explicit criteria is precision. A few habits quietly erase that advantage and leave you with something that behaves like a generic database again.
- Scraping everything a source will give you instead of applying the inclusion checklist — this just rebuilds a generic database one page at a time.
- Skipping domain-level de-duplication and sending near-identical outreach to two subsidiaries of the same parent company in the same week.
- Treating enrichment-tool estimates for revenue or headcount as exact figures in the email copy — vendors round aggressively, and a specific wrong number in a personalized line reads worse than a vague one.
- Mining once and reusing the list for months without re-verification, especially for signal-based criteria like 'recently hired' that expire by definition.
- Collecting more fields than the campaign will use — extra scraped data that never gets processed is just compliance liability with no upside.
Lawful Basis and How LDM Handles Sourced Data
Public availability of a business email or a filing doesn't remove the need for a lawful basis to process it. Under GDPR, processing business contact data for direct marketing typically relies on legitimate interest, which means documenting why the outreach is relevant to that specific company, offering a clear and working opt-out, and being able to show the data's origin if asked. Under CAN-SPAM, the requirements are more mechanical: accurate header and from-line information, a physical postal address in the email, and a functioning unsubscribe mechanism honored promptly.
Because a mined list is built from named sources against explicit criteria, it's easier to document than a bought database, where the provenance of any given row is usually unknown. Keep a simple source log next to the list: where each company came from, what date it was pulled, and what signal triggered its inclusion.
LDM's platform keeps this pairing intact end to end. Company records carry their enrichment source and the custom fields used to filter them, and campaigns pour from a defined list rather than an undifferentiated pool, so the criteria that got a company onto the list are still visible when someone reviews why it received an email.
FAQ
How is data mining different from buying a marketing database?
Buying a database means paying for someone else's aggregation, tagged broadly by industry and size, with unknown provenance and unknown freshness. Data mining means pulling companies yourself against specific criteria from sources you can check, which produces a smaller but far more current and defensible list.
What's a reasonable list size for a cold B2B campaign?
For most targeted campaigns, 300-800 tightly-matched companies per sourcing pass is a workable range. It's large enough to run a real campaign but small enough that a rep can keep personalization quality high across the whole list.
Which enrichment tool should I start with?
Start with whichever tool covers your primary firmographic gap, usually revenue or headcount confirmation, and treat it as a confirmation step after you've already narrowed candidates from a public or behavioral source, not as the primary discovery tool.
Is scraping public business data legal for cold outreach?
Accessing public records and job postings is generally lawful, but processing the resulting contact data for marketing still requires a lawful basis, such as legitimate interest under GDPR, plus compliance mechanics like CAN-SPAM's physical address and opt-out requirements. Public availability of a fact doesn't remove the obligation to handle it lawfully once it's used for outreach.
How often should a mined list be refreshed?
Re-verify anything older than 60-90 days before reuse, and refresh signal-based criteria like recent hiring or funding events sooner, since those signals expire faster than static firmographics like industry or headcount.
How do I avoid duplicate outreach across subsidiaries of the same company?
De-duplicate by domain rather than by company name when merging sources. Name matching misses subsidiary relationships and rebrands, which is how the same parent organization ends up getting near-identical emails from two different lists.
Want to apply this to your outreach?
We will map it to your segment and product — before any work starts.
Talk to us