The Complete Guide to B2B Contact Data (Full Course)

Find the right people at your target companies, get their emails in the cheapest order and verify each one. Example: paid steps drop from $55.00 to $32.96.

This is a full course on B2B contact data: how to find the right people at your target companies, get their email addresses, and check every address before you send.

We’ll go through one example step by step: Dapplecroft, an inventory tool for independent veterinary clinics in four US states (Colorado, Utah, Idaho and Montana).

Say the clinic list has 1,240 independent practices. The search finds 1,900 people with a buying title, and 1,400 of them end up with a valid email address. Removing duplicates and using the cheapest sources first cuts the cost of the paid steps from $55.00 to $32.96.

This guide is for founders, GTM (go-to-market) engineers and anyone who buys or builds contact data. It starts where the company list ends. If you already have emails and only want them clean, start at step 6.

1. Know what a contact record needs

A contact has four parts, and many files only have one of them. The email address gets all the attention. The other three decide whether the address is worth sending to.

  • Person: a full name you could look up. “Practice manager” is a job, not a person.
  • Role: what they do at this company today. “Owner” and “associate veterinarian” are different buyers at the same clinic.
  • Address: the mailbox, on the domain where the company receives mail (step 4).
  • Proof: where each of the other three came from. A link a colleague can open, or the tool and the day it answered.

Give every part a date. A job seen on a team page last week and the same job from a database record that’s two years old are not equally true. The file should show which is which.

For Dapplecroft, a complete row reads like this: Dr. Lena Ostrowe, owner, listed on brindlewickvet.example/our-team on December 3. Her address came from a database on December 4 and was verified the same day.

2. Find the people, flat-rate sources first

Start with the people, because an email address is only worth something if you know whose it is.

Pull every buyer, not one person per company. Write the buying titles down first. For Dapplecroft that’s owner, medical director, practice manager and hospital administrator. Then pull every person with one of those titles at every clinic. If the owner doesn’t reply, the practice manager might.

People databases. GetLeads and BlitzAPI are flat rate: one fixed price a month instead of a price per contact. Use them first. Prospeo goes last, because you pay per contact there. Apollo is another common option. We use BlitzAPI on its flat-rate plan.

The company’s own website. Small companies often list their team on an “about” or “team” page. For Dapplecroft this is the richest source, because clinics list their vets and their practice manager.

Public registers and associations. A state license lookup confirms that a vet is licensed. It almost never gives you an email. Association directories list members and board officers, often with addresses (step 9 shows the catch).

SourcePersonRoleEmailProof you can keep
People databasesIf they’re on LinkedInYes, can be out of dateOftenTool and date
The company’s websiteTeam and about pagesAs the company wrote itMostly shared inboxes like info@Page link and date
Public registersLicense holders, officersRarely currentAlmost neverRecord link and date
AssociationsMembers, board officersBoard titlesOftenPage link and date

Small local businesses are harder, because many owners have no LinkedIn profile. My TAM playbook for local businesses covers how to find those owners. This guide continues once you have names.

In the example, the databases and the clinic websites return 2,500 person rows with a buying title. They come from 1,240 independent practices. That count is realistic: the Census counts 1,688 veterinary offices in Colorado, Utah, Idaho and Montana (2023), including clinics owned by national chains and vets who only treat farm animals.

3. Remove duplicates before every paid step

Every tool that charges per row charges for duplicates too. So before each paid step, remove duplicates by the thing the tool charges for.

  • Email finders charge per person found. Remove duplicate pairs of name and domain.
  • Verifiers charge per address checked. Remove duplicate mailboxes (step 8).

In the example, the 2,500 person rows are 1,900 different people. Clinic groups list the same owner and manager on every location page. A group with six clinics shows the same practice manager six times. Without this step, a paid finder finds and charges for each copy.

Remove the people you shouldn’t contact before you pay for anything. Take out your current customers, open deals and everyone who asked not to be contacted, matched on the website domain against your CRM. Every email you buy for them is wasted.

Count before every upload. Rows in the file against unique people or unique mailboxes. The difference is money spent on answers you already have.

4. Check which domain receives the mail

The domain on the website isn’t always the domain that receives the mail. Four cases come up again and again:

  • Acquisitions. A group buys a clinic, keeps the local website and moves the staff onto the group’s email domain.
  • Rebrands. A new website domain, with email still on the old one, or the other way round.
  • Free email. The site has its own domain, and the whole practice reads one Gmail inbox.
  • No email at all. Some domains publish a record that says they accept no mail.

In the example, Cobblefen Pet Clinic runs its website on cobblefenpets.example. It joined Larchmere Veterinary Partners, and every address on its contact page ends in @larchmerevets.example. So the email finders have to run on larchmerevets.example, not on the website’s domain.

How to check. Look up the domain’s MX record, the DNS entry that names the server receiving its mail. Then compare it with the addresses the company prints on its own pages. A missing MX record proves little on its own. The addresses on the site are the better evidence.

The prompt I’d give Claude Code for this:

For every domain in domains.csv:
1. Look up the MX records (dig +short MX) and save the raw answer.
2. Label the domain: has_mx, null_mx (a single "0 ." record) or no_mx.
3. Open the homepage and the contact page. Collect every email address
   in mailto links and visible text, and count addresses per domain.
4. Set mail_domain to the domain most staff addresses on the site use.
   If the site shows no addresses, mail_domain = website domain.
5. Flag every row where mail_domain differs from the website domain,
   with the page link that shows it, for a human to check.
Run the email finders on mail_domain, never on a domain labeled null_mx.

5. Find the emails in the cheapest order

An email waterfall asks one source, sends only the people it missed to the next source, and stops at the first address that passes verification. The order decides what you pay.

The order I’d use:

  1. Emails that come with the people. GetLeads and BlitzAPI return emails for many of the people they know, with no cost per row. At BlitzAPI, the plan you pick decides whether emails are included. Verify each one (step 6).
  2. Then guess three formats and verify them. Most companies build their addresses the same way. For Lena Ostrowe at brindlewickvet.example, the address is usually lena.ostrowe@, lena@ or lostrowe@. Build these three, check them in that order with an email verifier, and keep the first one that comes back valid. If you already know one valid address on the domain, try its format first.
  3. Then a paid email finder for the rest. Prospeo and LeadMagic both charge per email they find, and nothing when they find none. Prospeo also doesn’t charge again when you look up the same person within 90 days.

Two rules for the order:

  • An address that isn’t valid counts as a miss. The person moves to the next source. Keep any catch-all address you got on the way, in case nothing better turns up (step 6).
  • A later source never replaces an earlier valid address. If two sources return two different addresses for one person, keep both with the source name and verify both.

Guessing doesn’t work on catch-all domains. A catch-all mail server accepts mail for any address, so all three guesses come back catch-all and you can’t tell which one is right. Send those people to the paid finder.

One more limit: a clinic with two vets called Sam can’t use first@ for both. When a guessed address fails, that person goes to the paid finder like any other miss.

In the example, the flat-rate databases return an email for 820 of the 1,900 people, and 640 of them verify as valid. The three guessed formats give a valid address to 400 of the remaining 1,260. That leaves 860 people for the paid finder. It returns 380 addresses, and 170 of them verify as valid.

Here’s the prompt I’d give Claude Code:

Build tools/emails.py. Read people.csv (one row per person, after
removing duplicates and do-not-contact rows) with first_name,
last_name and mail_domain.
Step 1: take the email from the database columns if there is one.
Step 2: for everyone without a valid address, build first.last@,
first@ and flast@ on mail_domain and verify them in that order.
Stop at the first valid one.
Step 3: send everyone still without a valid address to <finder API>
and verify what it returns.
Save every address with source, verifier result and date. Never
overwrite a valid address. Print people in, valid per step, catch-all
only, and nothing found.

Test the order on people you already know. Take 100 contacts from your CRM who have replied to you, so you know their address works. Run them through each source. Count the hits and the wrong hits, meaning addresses that verify but aren’t the one you know. Then divide what you paid by the correct hits. Order the sources by that number.

6. Verify every address and sort the results into four groups

A verifier answers one question: will this mail server accept mail for this mailbox right now? It asks the server without sending an email. It doesn’t tell you whether the person is the right one, or whether they’ll still be there next month.

MillionVerifier is the common pick. It doesn’t charge for catch-all and unknown results. Sort every result into four groups:

  • Valid: the server accepted the mailbox. The address goes into the campaign, with the date.
  • Invalid: the mailbox doesn’t exist, or the address is a temporary one. The person moves to the next source in step 5.
  • Catch-all: the server accepts mail for any address, so a normal check can’t tell a real mailbox from a wrong one.
  • Unknown: the check got no clear answer.

Drop “unknown” results. “Unknown” means the verifier couldn’t confirm the mailbox. Only send to addresses that came back valid, and to catch-all addresses that a catch-all tool cleared.

Then check the catch-all addresses with a second tool built for them, like LeadMagic or BounceBan. Use the cheap verifier for everything first and switch to the catch-all tool only for this group. They try to check each mailbox on its own. LeadMagic doesn’t charge when the answer stays unknown. Send email only to the catch-all addresses that come back valid.

These are some of the most valuable addresses on your list. Most senders drop catch-all addresses because a normal verifier can’t confirm them. So the people behind them get far fewer cold emails, and yours stands out.

Every verifier names its results a little differently. Map each label onto your four groups once, in code:

# One mapping for every verifier you use. Add each new tool's labels here.
GROUP = {
    "ok": "valid", "valid": "valid",
    "invalid": "invalid", "disposable": "invalid",
    "catch_all": "catch_all", "catch-all": "catch_all",
    "unknown": "unknown",
}

def group(label):
    return GROUP[label.strip().lower()]  # a label that isn't mapped stops the run

When a tool returns a label you haven’t mapped, the run stops and you decide what it means. That’s on purpose, so a default never decides for you.

In the example, the waterfall ends with 1,210 people who have a valid address. 300 people have only catch-all addresses. The other 390 have nothing usable, only invalid or unknown results. The catch-all check settles 240 of the 300: 190 valid and 50 invalid. 60 stay unresolved and are dropped.

The result: 1,400 people with a valid address and 500 without an email.

7. See what the order saves

Here’s what the order and the duplicate removal are worth in the example. Say a found email costs $0.04, a verification $0.004 and a catch-all check $0.01. Each approach starts from the same 2,500 person rows.

ApproachFinder lookupsPaid findsVerificationsCost
Every row to the paid finder2,5001,250 (300 of them repeats)1,250$55.00
Duplicates removed first1,900950950$41.80
Database emails, guessed formats, finder last8603804,440$32.96

The third approach costs 40% less than the first. It runs far more verifications, because each guess is one check, and a check costs a tenth of a find. It also reaches more people: 1,210 valid addresses before the catch-all check, against 950 found addresses in the second approach that still have to pass verification.

I left the flat-rate plan out of the table, because you pay for it every month anyway. If you buy one only for this list, add one month of it.

Checking the catch-alls pays too. The catch-all check of the 300 costs $3.00 at these prices and adds 190 valid addresses. Drop every catch-all instead, and you’d email 1,210 people instead of 1,400.

8. Treat different spellings of one mailbox as one

Remove duplicates by mailbox, not by the exact text of the address. Two addresses that look different can land in the same inbox, and then that person gets your email twice.

What the email providers say:

  • Gmail ignores dots. Dotted versions of a gmail.com address all reach the same inbox.
  • Except on work and school accounts. On a company domain that uses Gmail, dots do change the address. Never remove them there.
  • googlemail.com is the same as gmail.com.
  • Plus tags reach the main mailbox. janedoe+school@gmail.com arrives at janedoe@gmail.com. Microsoft 365 business mailboxes (Exchange Online) accept plus addresses by default too.
  • Capital letters don’t matter in practice. Lowercase every address before you compare.

In the example, one small practice runs everything out of a single Gmail inbox. Its contact page shows the address with a dot, a database has it without the dot, and an old directory has it with “+clinic”. Compare the text, and that’s three rows and three emails into one inbox.

The function I’d use everywhere:

GMAIL = {"gmail.com", "googlemail.com"}

def mailbox_key(address):
    local, _, domain = address.strip().lower().rpartition("@")
    local = local.split("+", 1)[0]       # plus tags reach the same mailbox
    if domain in GMAIL:
        local = local.replace(".", "")   # dots only don't matter on Gmail itself
        domain = "gmail.com"
    return f"{local}@{domain}"

Compare with the key, but send to the address as it was written. Use this one function for every comparison: removing duplicates, removing customers and opt-outs, and checking that no mailbox sits in two live campaigns. If two scripts use two different keys, they disagree about which addresses are the same, and no error message tells you.

9. Check whose address it is before you use it

An address on a page belongs to the person or company listed next to it, not to whoever owns the page. A scraper only knows which company owns the page, so the pipeline needs a rule for this.

Where it happens:

  • Association member directories. Every address there belongs to a member, not to the association.
  • Partner pages. A clinic’s “community partners” page lists a groomer, a pet sitter and a rescue, often with their Gmail addresses.
  • Manufacturer dealer lists. The addresses belong to the dealers.
  • Job posts. The address on a job post is often a recruiter’s.

In the example, Brindlewick’s partner page lists three Gmail addresses. They belong to a groomer, a pet sitter and a rescue, so none of them goes under Brindlewick.

The rule I’d write into the pipeline. An address counts for company X only if it’s on X’s mail domain, next to the name of a person you’ve already tied to X, or on X’s own contact page. A page that lists addresses from many different domains is a directory, whatever its title says. Addresses found there go to the company listed beside them.

One exception is worth keeping. Board officers on an association page often own a practice themselves. Match them to their own practice, and they’re some of the best rows you have.

10. Write down where and when you found every address

Fill these columns at the moment you find something. You can’t rebuild them later.

ColumnExample
source_typecompany site, database, register, directory, pattern
source_urlbrindlewickvet.example/our-team
seen_on2026-12-03
methodfound by a database, built from first.last, listed on a page
raw_labelthe verifier’s own word, for example catch_all
groupvalid, invalid, catch_all, unknown
verified_on2026-12-04
belongs_toown page, or third-party page plus the company listed beside it

With these columns you can re-verify only the old rows instead of everything. When a group of addresses starts bouncing, you can see whether they were found or built.

They also answer “where did you get my email?” in one lookup. If you email people in the EU, Article 14 of the GDPR asks you to tell them where their data came from, at the latest in your first message.

Then check again before every campaign. Every address and every job starts to age the day you check it. People leave or change jobs, clinics join groups and move to the group’s email domain.

Two public numbers show how fast this happens. In August 2026, 3.2% of all US employees left or lost their job in that single month. And the median time US employees had been with their current employer was 4.1 years in January 2026.

Neither one is a decay rate for your list. They do tell you that a file you don’t touch for three months is out of date in many places.

My defaults:

  • Re-verify every address older than 30 days right before it goes into a campaign. A check costs a fraction of a find, and the date column tells you which rows to check.
  • Check the proof again too. If the team page no longer lists the person, the address may still verify and reach the wrong person.
  • Add every bounce to the file as invalid, with the date, so no later campaign uses that address again.
  • When a whole domain starts bouncing, look for an acquisition before you blame the data source.

11. Run these checks before you send

These run on the final file every time, and one failed check stops the export.

  • Every row has a person, a role, an address and proof, each with a date.
  • The address is on the company’s mail domain, or the exception is written down.
  • Only valid addresses go into the email campaign, including catch-alls the second tool confirmed. Unresolved catch-alls and unknowns are dropped.
  • Every address was verified in the last 30 days.
  • No mailbox key appears twice in this file or in any other live campaign.
  • Customers, open deals, past opt-outs and past bounces are removed by mailbox key.
  • One person, one address. Keep the company address over a free email address.
  • No address from a third-party page without the company it belongs to.
  • Shared inboxes like info@ only where you meant to use them, with the person’s name kept for the greeting.

Then read ten rows by hand, picked from each group, and open the proof link for each one. It’s the only check here that can catch a row that passes every rule and is still wrong.

If a row is wrong, fix the rule that let it through and pick ten new rows. Stop when three rounds of ten in a row come back clean.

Where to start today

You can fix your next contact file today in three steps:

  1. Remove duplicate people and mailboxes before any paid step.
  2. Find the emails in the cheapest order: flat-rate databases, then guessed formats with a verifier, then a paid finder.
  3. Verify every address and write down where and when you found it. Only valid addresses go out.

Blog

Keep Reading

Start With a System Check

Start with a free target market map or a short system check. Either way, you leave with something useful.