Define the account by what happens inside the buyer’s sites: process type, NAICS code, plant count, installed equipment, permits and capex plans. Pull candidates from public records such as Census, SEC, EPA and federal award data. Score them into three tiers, map the buying committee at each tier 1 account, and verify every contact before you send.
What should define a target account besides headcount?
Physical and technical facts decide fit for an industrial or enterprise buyer. Headcount shows whether a company can pay, while process, sites and equipment show whether your robot, sensor, chip or software fits. Write the definition as filters a stranger could apply without asking you a question.
| Field | What to record | Where to check it |
|---|---|---|
| Process type | The exact process your product touches, such as palletizing, wafer inspection or wastewater treatment | Company site, job posts, permits |
| NAICS code | The six-digit code for the sites you sell into | Census, EPA and SAM.gov records |
| Plant or site count | Number of sites that run the process | 10-K property section, EPA facility list |
| Installed equipment | Machines, controllers or software already in place | Job posts, case studies, trade press |
| Regulatory status | Permits, inspections, recent violations | EPA ECHO, OSHA establishment search |
| Capex plans | Plant expansions, new lines, equipment budgets | 10-K, earnings calls, local permit filings |
“Mid-size manufacturers” fails the stranger test. “US meat processors in NAICS 3116 with two or more plants, a wastewater discharge permit and a 10-K that mentions automation spend” passes it, and every term in it maps to a public record.
Which public sources list industrial and enterprise buyers?
Census data sizes a segment, and facility, filing and award records supply the company names inside it.
- Census County Business Patterns. The Census Bureau’s CBP page, revised in August 2026, gives establishments, employment and payroll by industry, county and employment size, with 2023 as the latest data year. Use it to size a NAICS segment before you collect a single name.
- Economic Census. As of October 2026, the Census Bureau runs it every 5 years, for years ending in 2 or 7, and the 2022 cycle covers over eight million business locations and over 7,900 goods and services products. Product-level shipments show which segments make the goods your buyer uses.
- SEC EDGAR. EDGAR full-text search, checked in October 2026, covers electronic filings since 2001. Search 10-Ks for your equipment category next to phrases such as “new facility” or “capital expenditures” to find public buyers with budgeted projects.
- EPA ECHO and FRS. EPA states in October 2026 that ECHO covers more than one million regulated facilities. Its facility search shows the past five years of inspection and enforcement data, with names, addresses and permits. The FRS geospatial download adds latitude and longitude, so you can count sites per parent company.
- OSHA establishment search. It lists inspected workplaces by establishment name and state. Inspection history shows which plants run hazardous manual work that automation can replace.
- FAA and FCC licence records. For drone, aviation, satellite and radio products, the FAA aircraft registry and the FCC Universal Licensing System name the licence holders.
- SAM.gov and USAspending. SAM.gov holds entity registrations, contract opportunities and award data. USAspending, per its October 2026 about page, publishes federal contract, grant and loan awards above $25,000, as the 2006 FFATA law requires. Search both by NAICS and agency to find government buyers and the primes that serve them.
- Trade association member lists and conference exhibitor lists. These name the companies that pay to be in your category each year.
- Job posts that name equipment. A post for a maintenance technician on a named robot brand or PLC family tells you the installed base at that site.
How do you tier the accounts?
Score every account on six criteria from 0 to 2 points each, then cut the list at fixed thresholds. Fixed thresholds stop the team from promoting favourite logos.
| Criterion | 0 points | 1 point | 2 points |
|---|---|---|---|
| Process fit | Does not run the process | Runs an adjacent process | Runs the exact process |
| Site count | 1 site | 2 to 5 sites | 6 or more sites |
| Installed equipment | Unknown | Legacy or rival gear named | Gear you replace or integrate with, named in a post or filing |
| Regulatory pressure | None found | Permit on file | Recent inspection, violation or new rule on your problem |
| Capex signal | None | Expansion mentioned | New line, plant or equipment budget named in a filing or release |
| Reachable committee | No names | 1 or 2 verified contacts | 3 or more verified contacts across functions |
Tier 1 scores 9 to 12. The founder researches each one and writes each first message by hand. Tier 2 scores 6 to 8 and gets a sequence written for its segment. Tier 3 scores 0 to 5 and stays on a watch list until a new signal raises its score. If tier 1 holds more accounts than the founder can research in a week, raise the cut to 10.
How do you find the buying committee without naming errors?
Map at least one person in each function that signs off, then check that each record belongs to the right company. Gartner reported in May 2025 that buying groups range from five to 16 people across as many as four functions. In the same survey of 632 B2B buyers, run in August and September 2024, 74% of buyer teams showed unhealthy conflict. At an industrial account those functions are usually operations or the plant, engineering, EHS or quality, and procurement or finance. OT security joins when your product touches the plant network.
Contact databases make two errors that hurt a short list most.
- Namesake companies. A hiring signal for one company matched a different firm with the same one-word name, and a second signal matched an unrelated company with a similar name (our data, Oct 2026). Match on the website domain first. Take the top exact-name match across all company sizes, then apply your size filter.
- Wrong titles. One record at a data infrastructure company carried a co-founder title from a different startup (our data, Oct 2026). Drop any title that names another company, and confirm current roles on the company site or a recent filing.
How do you run the whole process in one week?
- Day 1. Write the account definition with the six fields above. Pull CBP counts for your NAICS codes to size the segment.
- Day 2. Export candidate facilities from ECHO, SAM.gov or USAspending, and add association and exhibitor lists. Roll sites up to parent companies.
- Day 3. Search EDGAR and job posts for equipment and capex terms. Fill in the installed equipment and capex columns.
- Day 4. Score every account and cut it into tiers. Read the tier 1 accounts by hand and fix any score you disagree with.
- Day 5. Map the buying committee for tier 1 and tier 2. Match contacts on domain, drop mismatched titles, and verify every email.
How often should you rebuild the list?
Re-score accounts every quarter and re-verify contacts within the week you send. When we re-checked 520 B2B contacts marked deliverable about 10 weeks earlier, 23% were invalid and 10% had turned catch-all (our data, Oct 2026). The full test is in how fast verified B2B email lists decay. With a few hundred buyers in a segment, each dead address costs you a real account, which changes the maths in whether outbound is worth it for deep tech.
To test your account definition and tiers against your pipeline before you send, book The Diagnostic.