Search “foundry” and you get hair salons: NAICS codes for B2B prospecting

Sunny Ray · May 26, 2026 · 5 min read

Search your prospecting tool for “foundry” and you'll get metal foundries. You'll also get digital foundries, alcohol foundries and hair foundries. Industry codes get you metal.

We learned this on a kickoff call on May 26 with an industrial software client. Their buyers are the companies that make and move metal. Charles was building the list live on the call, and the first pass on keywords came back polluted with design studios, distilleries and salons that happen to use the word in their name.

The hair foundry problem

Keyword filters match words, not businesses. “Foundry” is a fashionable word. Agencies use it. Breweries and distilleries use it. Beauty brands use it. A list built on the word will look big and feel productive right up until your sequence starts telling a hair salon about casting defects.

That doesn't just waste sends. Every irrelevant email is a chance for a spam complaint, and every complaint hurts the sending domains that need to reach the real buyers. A dirty list is a deliverability problem wearing a data costume.

Broad industry tags in the big databases didn't save us either. In my experience industrial businesses are especially easy to misfile under a generic label. You need a classification built for the job.

Sizing the ICP: three markets of a few thousand companies

Before we pulled a single contact, we sized the three ICPs the client cared about. These were the working estimates on the call:

• Foundries: about 1,800 to 2,200 companies in the US.
• Metal service centers: about 6,500.
• Steel manufacturers: about 1,500 to 1,800.

Those numbers change how you run the campaign. A market of 2,000 companies isn't a list you can burn through and replace. Every account matters, so precision beats volume. You'd rather send to 1,500 real foundries than 5,000 records where a third are wrong.

NAICS codes plus AI targeting

So the list moved off keywords and onto NAICS codes, the North American Industry Classification System, combined with AI targeting to check each company.

For foundries, the codes are specific. The foundry group, 3315, breaks down into six-digit codes like 331511 for iron foundries, 331512 for steel investment foundries, 331513 for steel foundries other than investment, and separate codes for aluminum, nonferrous die-casting and other nonferrous foundries. Metal service centers sit in 423510, metal service centers and other metal merchant wholesalers. No hair salon is going to file under any of those.

Vendor guidance found in search says the same thing we saw: build filters at the six-digit level, and watch for misclassified companies anyway. Codes get you most of the way. The AI targeting is there to catch what's left, so the list holds companies that actually pour or move metal.

We combined three sources to build it: Clay, Sales Navigator and Apollo, each filtered on the codes. No single database has every industrial company with clean data, and these are businesses that often have thin web presence. Overlapping sources fill the gaps.

Seed accounts and the do-not-contact list

Two inputs from the client made the list much better than anything we could build alone.

Seed accounts. The client sent examples of ideal customers through email and a shared Slack group, including a set of 52 companies in the 10 to 50 million revenue range. Seeds give the AI a picture of “more like these,” which beats any written ICP description.

A do-not-contact list. Existing clients and deep-pipeline prospects came out before anything went out. Nothing kills a sales relationship faster than a cold sequence landing on a customer you already have, or on a prospect your team is three meetings into.

Launch the same day, review the first list together

Speed mattered here. The client's sending infrastructure was ready: 100 mailboxes with warm-up nearly complete. So the plan was to launch the first campaign, aimed at the foundries ICP, that same day, as soon as the client reviewed and approved the target list.

That review is the most underrated step in list building. The client knows their market in a way no filter does. A quick scan of the first list catches the edge cases: a company that's technically a foundry but too small, a competitor, a distributor that looks like a manufacturer. Approve the list together and you also share ownership of the results.

One segment, one campaign

The engagement was set up for four campaigns a month. With three ICPs, that means each segment gets its own campaign and its own copy, instead of one generic message stretched across foundries, service centers and steel makers. A service center that buys and moves metal doesn't have the same day as a foundry pouring it, and an email that pretends otherwise reads like it was written for nobody.

Foundries went first because the list was ready first. The other two follow once their lists pass the same review. Updates run through the shared Slack group and WhatsApp, so the client sees what's going out and what's coming back without waiting for a weekly call.

The method for any industrial ICP

• Size each segment before you pull contacts, so you know how precious each account is.
• Filter on six-digit NAICS codes, not keywords.
• Run an AI check over the result to catch misclassified companies.
• Combine at least two or three data sources.
• Seed with the client's best customers and strip a do-not-contact list.
• Have the client approve the first list before launch.

Industrial buyers are some of the most reachable people in B2B, because almost nobody emails them well. A clean list is most of the advantage. If you want to see how we build them, here's how we work.

Frequently asked questions