Buy, Scrape or Build: Where B2B Lead Data Actually Comes From

Every outbound campaign runs on B2B lead data, and almost every campaign that fails quietly blames the copy when the real culprit was the list. There are only three ways to get contact data: buy it, extract it, or build it yourself. Each has a different cost curve, a different accuracy profile and a different failure mode. Knowing which one you are actually using – and what it does badly – is the difference between a 4% reply rate and a burned domain.

Option 1: Buying data from a provider

The database model. You pay for seat access or credits and export contacts filtered by title, industry, headcount, geography and sometimes technology or intent signals.

What it is good at: speed and breadth. You can go from nothing to 5,000 targeted contacts in an afternoon, with firmographics attached, and the coverage of mid-market and enterprise companies in English-speaking markets is generally solid.

Where it breaks:

• Staleness. Providers refresh on a cycle, not continuously. Roughly a fifth to a quarter of B2B contact records go bad each year through job changes alone, so any export includes a layer of records that were true nine months ago.
• Shared targeting. If you filter on obvious criteria, so does everyone else buying from the same provider. Your prospects are receiving eleven other emails from the same slice of the same database.
• Guessed emails. Many providers pattern-match rather than verify. A record marked “high confidence” is often just first.last@domain with no delivery evidence behind it.
• Thin small-business coverage. Local services, trades, independent clinics and family-run operations are poorly represented almost everywhere.

Practical rule: never send to a purchased export directly. Run it through email verification first and expect to discard 10-25% of what you paid for. That discard rate is the real price of the data.

Option 2: Extracting data from live sources

Here you pull from places where companies publish information about themselves: professional networks, association directories, licensing boards, job boards, review platforms, government registries, event exhibitor lists, procurement portals.

What it is good at: freshness and differentiation. A directory of licensed contractors updated last week beats a database snapshot from last quarter, and nobody else in your category has bothered to compile it. Extracted sources also carry natural qualification signals – a company posting three sales roles is behaving differently from one that is not.

Where it breaks:

• Company-level, not person-level. Most public sources give you an organisation and a generic inbox, not a named decision-maker. You still have to find the human.
• Maintenance cost. Sources change structure, add friction, or restrict access. What worked last quarter may need rebuilding.
• Terms and legality. Public visibility is not the same as permission to collect and store. Check the source’s terms, and check your obligations under cold email compliance rules before the list ever gets loaded.
• Messy output. Inconsistent naming, duplicated entities, missing domains. Expect real cleanup work.

This is usually the strongest option for niche verticals, and it is why industry-specific lists – the kind used in manufacturing list building or trade-specific outreach – outperform generic database pulls so consistently.

Option 3: Building your own list from scratch

Manual or semi-manual research: define the account universe first, then find the right person at each account, then find and verify their contact details. Slow, and often the highest-yielding option for high-value targets.

A workable sequence:

1. Define the account universe. Write firmographic criteria and explicit disqualifiers. If you cannot say what makes a company a bad fit, your list will be full of them.
2. Enumerate accounts. Pull the company layer from whichever source covers your niche best – it is far easier to get accurate company lists than accurate contact lists.
3. Identify the right role per account. Not the most senior person; the person whose weekly problems your offer solves. In a 40-person company that is often two levels below where people instinctively aim.
4. Resolve and verify contact details. Find the address, then confirm it independently.
5. Add the context field. One specific, true detail per record – a recent hire, a new location, a stated priority. Without it, personalisation collapses into mail-merge.

Where it breaks: throughput and consistency. Fifteen to forty accounts per researcher-hour is realistic for genuinely researched records, and quality drifts unless you enforce a field standard. Full detail on the workflow lives in our B2B list building guide.

The hybrid approach most good teams actually use

In practice the three options are layers, not alternatives:

• Use a purchased database to establish the account universe cheaply and fast.
• Use extracted public sources to add the accounts the database missed and to attach timing signals.
• Use manual building only on your top tier – the accounts worth 20 minutes of research each.
• Run everything through data enrichment and verification before it touches a sending tool.

Segment by value. A $60k-ACV target deserves hand-built research. A $400/month target does not, and pretending otherwise is how teams end up spending research hours on accounts that can never pay for them.

How to judge any data source in one week

Do not evaluate providers on their marketing claims. Run a controlled test on 200 records:

1. Verification pass rate. What percentage survives verification? Below 85% on a paid source is weak.
2. Title accuracy. Manually check 25 records against the company’s own site or the person’s public profile. Count how many titles are current.
3. Fit rate. What percentage genuinely match your ICP rather than merely matching your filter?
4. Bounce rate on a live send. The only test that fully counts. Above 3% and something upstream is wrong.
5. Cost per booked meeting. Not cost per record. A source at three times the price per record can still be cheaper per meeting.

The failure mode to avoid

The expensive mistake is not choosing the wrong source. It is treating any source as finished data. Every list – bought, extracted or built – is a hypothesis about who exists and what they care about, and it decays from the moment you create it. Build verification and refresh into your process as recurring work, not a one-off step, and the source you chose matters far less than the discipline you apply to it.