How to Build a B2B Outbound List That Is Not 80% Wrong

Picture a founder who buys a list of 4,000 "logistics companies" and starts sending. By week 2 the replies tell the story: a software vendor that writes about logistics, a university research group, a moving company, two of the founder's own customers, and a lot of bounces.

None of those are deliverability problems or copy problems. They are list problems, and they were decided before the first email was written.

This article covers how to build an outbound list that holds up: the three gates every contact must pass, where the data comes from, why filtering by meaning beats filtering by keyword, how to verify and refresh before every send, and how big the list should be.

Why Outbound Lists Go Wrong

Many lists are built in one pass: pick an industry tag, a headcount range, a location and a few job titles in a data tool, then export. Every filter in that pass is coarse.

• Industry tags are broad. A single tag can cover manufacturers, distributors, software vendors and consultants in the same field.
• Company descriptions are ambiguous. A keyword like "freight" matches carriers, brokers, freight audit software and a blog about shipping.
• Job titles drift. "Head of Growth" means marketing at one company and sales at another.
• Contact data decays. People change jobs, and the email that was valid when the database last checked it may be dead today.
• Your own accounts sneak in. Without a check against your CRM, customers and open deals get cold emails meant for strangers.

Each error is small. Stacked together, they make a large share of a list wrong before anyone writes a word.

The Three Gates Every Contact Must Pass

A list is only as good as the filters between the raw pull and the send. We use three gates, in order, and a contact has to pass all three.

Fig. 1Three gates every contact must pass

1. Raw pull from data toolsExample: 5,000 companies

   Industry, size and location filters only.

2. Gate 1: company fitExample: 1,200

   Checked against what the company actually does.

3. Gate 2: persona fitExample: 900 accounts, 1,800 people

   The right roles found inside each company.

4. Gate 3: not already yoursExample: 800 accounts

   Customers, open deals and recent conversations removed.

5. Verified and approvedExample: 1,500 sendable contacts

   Emails verified on the day, list approved by a person.

1. Company fit

Does the company actually match your ICP? Not the industry tag, the business itself: what it sells, to whom, at what size, in which market. This is the gate that removes the software vendor from a list of carriers. If your ICP is not written down yet, start with How to Build an ICP Matrix: 1 ICP, 3 Segments, 2 Personas.

2. Persona fit

Inside each company that passes, find the people who own the problem you solve. Usually that is 1 or 2 roles per segment, matched on what the role does rather than the exact title. Check seniority too: a coordinator and a VP may share a department and have nothing in common as buyers.

3. Not already yours

Remove every account you already have a relationship with:

• Customers, including past customers who might be a win-back but deserve a different message.
• Open deals, which belong to whoever owns them.
• Recent conversations, such as anyone who replied, booked or was contacted in the last few months.

This gate depends on a clean CRM. If yours is not, Fix Your CRM Before You Scale Outbound covers the order we use.

Where the Data Comes From

No single source covers everything. A list is usually assembled from several, each doing a different job.

Order the sources by cost. Pulling companies from a database is cheap; reading websites, enriching contacts and verifying emails costs time or credits per record. So run the cheap filters first, apply company fit to what survives, and only enrich and verify the contacts that passed every gate. Enriching 5,000 contacts to send to 1,500 pays for 3,500 records you throw away.

Keep a record of where each field came from and when. When a reply says "I left that company in the spring", you want to know which source supplied the stale job title, so you can trust it less next time.

Clay is a common way to chain these sources together, but the tool matters less than the order of the gates. Signals from these sources also decide timing, which is covered in Lead Scoring That Works: Fit, Timing and Signal Decay.

Filtering by Meaning, Not Keywords

Gate 1 is where many lists fail, because it is often done with keywords. A keyword filter asks whether a company's description contains a word. It cannot tell which meaning of the word the company uses.

Here is one system our founding team built where this mattered.

The constraint. An immigration services firm needed a lead discovery pipeline. The trouble was its name: a common word that appears in thousands of unrelated contexts. Keyword matching brought back mostly noise.

What we built. Semantic filtering: instead of matching the word, the pipeline judged whether each result was about the right thing, using the surrounding context. It processed 12 months of historical data and then ran as weekly automated monitoring.

The result:

• False positives went from 80% to 10%
• Qualification time went from days to hours
• Weekly automated monitoring

The same approach works for company fit in any list. Here is how the two methods compare.

Fig. 2Keyword filtering vs semantic filtering

Keyword filterSemantic filter

How it decidesDoes the text contain the wordDoes the company match the meaning

Common wordsLets every other meaning throughSeparates meanings by context

SetupMinutesA labelled sample and testing

Review loadHigh, many false positivesLow, mostly edge cases

Best forRare, specific termsBroad or ambiguous categories

In practice, semantic filtering means giving a model the company's own description and a short, precise definition of your ICP, then asking for a fit judgement with a reason. Start with a labelled sample of 50 to 100 companies you have already sorted by hand, check the model's calls against yours, and tighten the definition until they agree. Then run the full list, and have a person spot-check the edge cases.

Verify and Refresh Before Every Send

A list that was right when it was built can be wrong by the time it is sent. Contact data decays every week as people change roles and companies close inboxes.

The rule we follow: verify right before every send, not once a quarter.

• Email verification on the day. Every address checked shortly before it goes into a sequence. Hard invalids removed, catch-all addresses sent at lower volume or held back.
• Job and employer check. Anyone whose title or company changed since the pull is removed or moved to the right list. A job change can itself be a signal, which is covered in Executive Movement Tracking.
• CRM re-check. Gate 3 runs again, because a contact may have booked a call or entered a deal since the list was approved.
• Suppression list. Anyone who unsubscribed or asked not to be contacted is removed from every future list, permanently.

This matters for more than reply rates. Bounces and spam complaints hurt your sending domains, and since February 2024 Gmail and Yahoo require bulk senders to keep complaints below 0.3%. The contact data refresh play shows how to run this check automatically before each send.

How Big Should the List Be

List size should come from your sending capacity and your segments, not from how many rows a data tool will export.

Start with capacity. A warmed cold inbox sends about 40 emails a day, so volume comes from more inboxes, not more emails per inbox. Our own setup runs 12 inboxes across 5 sending domains, about 480 emails a day at full warmup. The math for your setup is in How Many Domains and Inboxes Do You Need for Cold Email?

Then do the arithmetic. As an example: if each contact gets 3 emails in a sequence, 480 sends a day supports roughly 160 new contacts a day once follow-ups are running. Over 20 sending days, that is about 3,200 new contacts a month.

Then split by segment. An ICP matrix of 3 segments by 2 personas gives 6 lists, each with 3 angles to test. Each list needs enough contacts to tell the angles apart, which is usually a few hundred accounts, not thousands.

Sizing also sets the refresh rhythm. At the example pace above, a list of 3,200 contacts is used up in a month, so the next list has to be built, filtered and approved before the current one runs out. Plan list building as a weekly job, not a one-off project. A team that builds one big list in January and sends from it until June is sending to people whose data aged months ago.

Leave room for signals, too. A share of each week's capacity should go to accounts that just showed a buying signal, such as a new leader or a funding round, because those contacts are worth more this week than next month.

That is why a precise list of 300 accounts with live signals beats a loose list of 5,000. The smaller list teaches you something. The larger one mostly teaches your domains to land in spam.

Mistakes That Make a List Wrong

These are the common list mistakes:

• Trusting industry tags. A tag is a starting filter, never the final word on company fit.
• Filtering on titles alone. Matching "VP Sales" across every company size puts very different buyers on one list.
• Skipping the CRM check. Cold-emailing your own customer is the fastest way to look careless.
• Verifying once. A list verified at pull and sent 6 weeks later has decayed in between.
• Exporting everything. A list sized to the data tool rather than the sending plan burns contacts you could have reached with a better angle later.
• No human review. A person should approve every list before it sends. 10 minutes of spot-checking catches the patterns a filter missed.

FAQs

Build Your Outbound Lists With Oriel

Oriel is the engine room behind the go-to-market team at B2B companies above $100k/month in revenue. For list building, we turn your ICP into segment and persona lists, filter companies by what they actually do, remove everyone who is already yours, and verify every contact before it sends. You approve every list, and everything lives in a CRM you own.

On a 30-minute call, tell us who you sell to and how you build lists today. We will show you where the wrong contacts are likely coming from and what the first 6 lists would look like. If you are not ready for this yet, we will tell you that too.