How to Build a B2B Lead List: From ICP to Clean, Actionable Records

Define the account universe first

Start with companies, not contacts. Translate the ideal customer profile into observable account criteria such as industry, geography, employee range, business model, installed technology or growth stage. Add explicit exclusions for segments you cannot serve, even when they resemble good customers on one dimension.

Create a named account universe before searching for people. Starting from accounts keeps the list anchored to companies you can realistically sell to, rather than to people who merely happen to match a title.

Select personas and buying-group coverage

For each account, identify the roles involved in the problem: likely owner, daily user, technical evaluator, economic buyer and blocker. Job titles vary, so use department, seniority and responsibility—not a single exact title string—to define the persona.

Large accounts often require more than one contact. A list with one executive per company can be too brittle; a list with twenty loosely relevant contacts creates noise. Set a target number of useful buying-group contacts based on deal complexity.

Source and validate contact data

Use a database or research workflow to obtain work emails, phones and role information. Validate the fields that matter to your outreach channel before import. For database selection, compare B2B contact databases; if one source leaves important fields missing, use the data-enrichment guide to design a second pass.

Record the source and date for critical fields. A list built six months ago may still contain valid companies but stale titles, and a job change can turn a previously correct email into the wrong person for the account.

Deduplicate and normalize before CRM import

Deduplicate at both account and contact levels. Normalize domains, company names, country/state values and seniority labels so routing rules do not treat equivalent records as different entities. If the CRM already contains the company or person, update the intended fields rather than creating a parallel record.

Create an import contract: required fields, allowed overwrite fields, owner-assignment rules and suppression fields. This is faster to define once than to repair after every campaign.

Run list QA before outreach

Take a random sample and manually inspect it. Ask: Is this company truly in the ICP? Is the persona relevant? Is the contact still in role? Does the selected outreach channel have a usable address or number? Are any customers, partners or opt-outs present? Fix the pattern that produced each error rather than deleting only the bad row.

Once the list passes QA, connect it to a sequence or a founder-led manual outreach motion. Startups building their first motion can see how this fits Apollo in Apollo for startups.

Worked example: from ICP to import file

Suppose the ICP is US-based cybersecurity software companies with 50–500 employees, and the outreach targets heads of demand generation plus one marketing-operations contact. Build the account set first. Then attach personas. This prevents a list from becoming a random collection of individual contacts that happen to match job-title keywords.

Lead-list QA before any sequence starts

• Remove exact duplicate contacts by stable identifiers such as CRM ID, email or profile URL.
• Normalize company domains and names before deduplicating accounts.
• Check that every contact still works at the target company and holds a relevant role.
• Separate contacts with no usable channel from contacts that are valid but intentionally not being contacted.
• Apply existing customer, open opportunity, competitor and do-not-contact suppression lists.
• Tag source, build date and list logic so the list can be audited later.
• Import a small batch first and inspect field mapping before loading the rest.

The finished list should be explainable. A manager should be able to answer why each account is present, why each person was selected, which channel is allowed, and what should happen when the record changes. If enrichment is needed before import, use the data-enrichment guide; if the targeting rules are still vague, fix the ICP first.