How to Build a B2B List Without Paying for Bad Data
A lot of early GTM work gets stuck on "building the list," as if volume were the craft. I have sympathy for the opposite warning: new GTM engineers should not learn list building first. The first skill is knowing who belongs on a list at all. Data also decays — people change jobs, inboxes die — which is easy to forget if you bought a file last quarter and still treat it as fresh. And a point I want every founder to take seriously: give two "AI SDRs" the same messy CRM, and you will get the same wrong email. The software will not rescue bad rows. In consultants threads about automating outbound, validation keeps showing up as the unglamorous work people try to skip.
Pramod Kumar, co-founder of Kingdom of Kumar. On X
·
This is how I would build a small, trustworthy list, and how I would refuse to pay for a larger untrustworthy one. Enrichment order is in Enrichment waterfalls. Fit is in How to define an ICP your team will actually use. Once the list is in a system, CRM hygiene takes over. The surrounding craft is on GTM engineer and GTM engineering. If the offer is not ready, take the App Launch Score and do not buy rows.
"New GTM engineers shouldn't learn list building first"
I agree with the spirit. If you learn "how to extract 10,000 contacts" before you learn "who we are allowed to contact, and what must be true of a row," you will become an expensive way to fill sequencers with junk.
The first list I want is short and argued-over: companies and people who match a written fit rule, built by hand from conversations, research, and maybe one database search you actually read. Only then do you enrich, verify, and, if it still hurts, automate a slice. A junior or a first GTM engineer who is given a list-building quota without a fit rule will do damage, which is a theme I will come back to in the hygiene article.
Founders make the same mistake when they buy a file because a vendor showed a count. A count is not a list.
Emails decay, and so do titles
A work email is a perishable fact. People leave, catch-all domains accept mail that nobody reads, and a "verified" tag from last quarter is not a verification now. Titles move too. If you send to a stale role, you are not only being ignored — you may be training the new person to see you as careless.
I would recheck contact data before you reuse a list, especially after a pause, and I would store the date of last verification on the row. I will not invent a decay percentage. You can see your own bounce and "no longer here" rate if you look.
"Give two [AI SDRs] the same messy CRM and you will get the same wrong email"
Yes. Models and tools will confidently use whatever is in the field. If the field is wrong, personalization makes it worse: a first line that mentions the wrong company, the wrong city, the already-churned customer marked as a prospect.
Clean the record before it enters any drafting or sending tool. "We will let the AI SDR sort it out" is not a plan. Two tools on the same mess will only agree on the mess.
The practical implication: the CRM, or the sheet that is standing in for one, is part of list building. If "already spoken to," "customer," and "do not contact" are unreliable, stop adding new rows until those flags can be trusted on a sample you checked by hand.
A list-building method that stays small
1. Write the fit and never-contact rules first. No rows until that page exists.
2. Start from people you already know, or from a manual search you review line by line. Twenty to a hundred accounts is a good first range for a founder motion. I am not giving you a scientific number. I am telling you to be able to look at every row.
3. Enrich only the fields that gate sending. Usually: work email, role, company, one evidence field. See the waterfall article. Cap the spend.
4. Verify the email. Failed rows do not enter the sequencer.
5. Deduplicate against customers, open deals, recent outreach, and unsubscribes. If you cannot do this, you do not have a list. You have a new pile.
6. Date the list. A "built on" and "last verified" date belongs at the top. After a few weeks, verify again before you scale, or before you resume.
7. Review the bounces and "wrong person" replies on Friday, and fix the source, not only the row.
If a vendor or a freelancer skips steps 1, 4, or 5, they are selling you a file, not a list.
What I would not pay for
- A giant file with no dates and no method
- "Lookalike" rows you will not inspect
- Personal emails or mobiles you have no policy or reason to use
- A second database because the first one had gaps, before you measured the gaps
- Recurring enrichment on the whole file when you only contact a slice
Pay for sources and verification that improve the rows you will actually use this month. Everything else is inventory that decays on the shelf.
A half-day audit of a list you already have
Before you buy another source, take the file or CRM view you already send from and do this by hand, on a sample of fifty rows:
- How many are obviously the wrong fit?
- How many are customers, closed deals, or people you already emailed this quarter?
- How many emails bounce, or look like personal inboxes you have no policy for, when you verify them now?
- How many titles look stale against a public page you can actually see?
- How many rows lack a date of last check?
Write the counts in a Friday note. If the sample is messy, the fix is cleanup and a tighter rule, not a larger purchase. If the sample is clean, you may have earned a small expansion — still verified, still capped.
This audit is also how you brief a freelancer. "Make the list bigger" is a bad brief. "Here are the rules, here are the fifty-row results, here is the cap" is a brief you can supervise.
FAQ
Is buying a list ever all right?
If you will inspect it against your fit rule, verify, and dedupe, a purchased starting set can be an input. If you will upload and send, no.
How do I know my data is "good enough"?
You looked at a sample, verification is in place, bounce and wrong-person rates are something you watch, and the never-contact flags work. I will not give you a target percentage to hide behind.
Can you build the list for us?
Kingdom of Kumar builds and runs a scoped system on Grok and Claude for $500 a month, three-month minimum. You own it, and you approve what goes out. We will not dump a big unverified file into your sequencer to look busy. You will still own the fit rule. This is not a $135,000 data-ops hire, and it is not a list shop.