Installs, repairs, and maintains heating or cooling systems.
Uses HVAC throughout its description because contractors are its customers.
Mentions HVAC workflows without doing HVAC work.
Excludes companies whose business model shows they sell to the market.
Most teams start an Apollo search by writing down what they want. The stronger operator also writes down what must never enter the list.
This becomes critical when the target is 20,000 to 50,000 leads across a large US geography. Broad searches often depend on company keywords. Apollo says those keywords can come from structured tags and free-text fields such as the company description, website, and other profile data. That gives useful reach, but it also creates semantic overlap.
A home-insurance agency, an insurance lead-generation firm, and software built for insurance brokers may all mention the same phrases. Only one is the buyer.
The operating rule: an inclusion keyword finds a conversation. A negative ICP determines which side of that conversation the company is on.
Why broad Apollo keyword searches produce believable junk
A false positive rarely looks absurd. It usually has the right geography, company-size range, and words. The error lives in the business model.
Apollo's documentation warns that broad keywords can match companies that merely mention a term, such as a technology stack or partner market. Apollo recommends refining results with other filters and supports keyword exclusions, excluded industries, company lists, and precise SIC or NAICS classifications.
Step 1: define the positive and negative ICP together
Positive ICP What the company sells Who it sells to Operating model, geography, size, and ownership Observable website evidence Negative ICP Vendors serving the target Agencies marketing to the target Software built for the target Directories, recruiters, consultants, and educators Excluded parent, franchise, or ownership models
The negative ICP should describe categories, not just disliked companies. If you only block known domains, the next vendor with the same business model will pass.
Step 2: mine exclusion language from real false positives
Run a deliberately broad sample, then review company websites and descriptions. For every wrong company, record the phrase that revealed why it was wrong.
Useful vendor-language patterns often include “for contractors,” “serving agents,” “marketing solutions,” “software platform,” “lead generation,” “consulting,” “recruiting,” “training,” and “directory.” Do not paste this generic list into every campaign. Use false positives from the specific niche to create the final vocabulary.
Step 3: separate hard exclusions from review signals
A hard exclusion should fail the record automatically. Examples include a blocked industry, known competitor domain, prohibited territory, or a clear phrase such as “digital marketing for HVAC companies.”
A review signal should route the record to CHECK. A company may say “serving homeowners” because it performs residential work, or because it sells software to businesses that serve homeowners. The phrase alone is not enough.
Store four fields with every decision: company URL, decision, exclusion reason, and supporting evidence. That is how the same negative ICP can be reused without becoming a mysterious Boolean string.
Step 4: test the exclusion list on a difficult sample
Use the 20-profile Apollo acceptance test, but deliberately include vendors and adjacent businesses. Sample later pages, ambiguous descriptions, franchise locations, subsidiaries, and companies with multiple business lines.
After every review round, calculate false-positive rate by exclusion reason. If agencies keep appearing, the search needs another structural filter, not more optimistic reviewing.
Step 5: save the search and its exclusion version
Apollo says saved-search counts can change as its data changes. Record the date, included filters, excluded filters, sample results, and approval version. If somebody removes one negative keyword later, the team should be able to explain the difference.
Frequently asked questions
What is a negative ICP?
A negative ICP is the set of company types, business models, attributes, and evidence patterns that disqualify an account even when it shares words with the target market.
Can Apollo exclude company keywords?
Yes. Apollo's Industry & keywords filters support keyword and industry exclusions. You can also exclude specific companies or lists of domains.
Should I use keywords, industries, or SIC and NAICS codes?
Use them together when possible. Keywords capture niche language, industries provide broad structure, and SIC or NAICS codes can improve consistency.
How many negative keywords should I use?
Use enough to remove repeated false-positive clusters without excluding real buyers. Test every addition against known positive and negative examples.
AnswerThePublic research note
The English, United States report for Apollo negative keywords surfaced questions around improving targeting with negative keywords, negative search terms, and best practices for exclusions. It was generated on August 3, 2026.
Sources
Need the right accounts before you enrich?