Every outbound campaign lives or dies on one thing that most teams get wrong from the start: list quality. You can write the perfect email, set up flawless deliverability, and still get 1% reply rates — because you’re reaching the wrong companies.
Industry standard list building achieves 50–60% accuracy for niche ICPs. That means up to half your outreach budget is wasted on companies that will never buy. There’s a better method that costs nothing extra and hits 90–95% accuracy: filtering on LinkedIn’s “About Us” descriptions using semantic keywords. Here’s the full system.
Why Traditional List Building Fails for Niche ICPs
Most outbound teams build their lists using industry codes — SIC, NAICS, or the proprietary tags data providers assign. These codes have a fundamental flaw: they’re assigned by data providers based on outdated, often self-reported classifications. For broad markets, they’re good enough. For niche ICPs, they collapse.
A company that sells AI SDR software might be categorized as “Computer Software,” “SaaS,” or “Marketing Automation” — depending on which provider you use. A dental DSO might show up under “Healthcare,” “Medical Practices,” or “Hospital Administration.” None of these tags are specific enough to tell you whether a company matches your ICP.
The result: 40–60% of your list is wrong. You’re emailing competitors, companies too small or too large, or businesses in the wrong vertical — all while burning your sender reputation and your team’s time.
The LinkedIn About Us Advantage
LinkedIn’s “About Us” description is the most underused data point in B2B outbound. Every company writes it themselves — to attract customers, recruit talent, and communicate their positioning. That means it’s always:
- Current — Companies update it when their positioning changes
- Accurate — Self-reported, in their own words
- Specific — Uses exact category and product terminology
- Free — Available in standard data exports, no extra cost
Instead of relying on a data provider’s industry tag, you filter on the exact language your ideal customers use to describe themselves. That’s the difference between “Healthcare Industry” and “orhodontic,” “dental imaging,” “clear aligner practices.”
The Three-Step Semantic Filtering Process
Step 1: Enrich Your List with LinkedIn About Us Text
You need the About Us field pulled into your dataset. This is a standard enrichment available through:
- Clay.com — Recommended. Native LinkedIn company enrichment, the About Us field is a first-class data point
- Apollo.io — Export to Google Sheets, includes company description fields
- Custom scraper — For advanced users who need full control
Start with any broad list: LinkedIn Sales Navigator export, Apollo search, ZoomInfo download. What matters is having the LinkedIn company URL so you can pull the About Us text.
Step 2: Build Your Semantic Keyword Bank
This is where most of the value is created. Your keyword bank should reflect the exact language your ICP uses — not the language you use to describe them.
Manual method (most accurate, 30–60 minutes):
- Pick 3–5 dream accounts — companies you wish you could clone 10,000 times
- Open each company’s LinkedIn About Us section
- Manually extract high-signal terms: product names, category language, compliance terms, pain-point vocabulary
- Build your list — aim for 10–30 keywords
AI-assisted method (10–15 minutes, nearly as accurate):
- Copy/paste About Us text from 5 dream accounts into Claude or ChatGPT
- Prompt: “Extract every relevant keyword and category term that describes these companies. Comma-separated.”
- Review the output, remove generic terms, keep high-signal ones
What you’re looking for: terms that only your ideal customers would use. Generic terms like “B2B” or “enterprise software” are noise. Specific terms like “revenue cycle management,” “OEE tracking,” or “clear aligner practices” are signal.
Step 3: Build the Compound Filter
Combine industry buckets with your semantic keyword bank:
- Industry IN [5–10 relevant industry tags] — casts the wide net
- AND About Us CONTAINS [your 10–30 keywords] — applies the precision filter
The industry filter gets you into the neighborhood. The About Us keywords put you in front of the right house. Together they achieve 90–95% ICP accuracy versus the 50–60% you get from industry codes alone.
Validation: How to Know Your Filter Is Working
Before running your filter across your full TAM, validate it:
- Run the filter on a sample of 100 companies
- Manually review each — does the About Us match your ICP?
- Calculate accuracy: (Matches ICP / Total) × 100
- If accuracy is below 85%, tighten the keyword list or add negative filters
- Test that your known dream accounts pass the filter (check for false negatives)
This validation step takes 60–90 minutes but saves you from wasting an entire campaign on a bad list.
Advanced Techniques
Negative Filtering
Exclude terms that create false positives:
About Us CONTAINS [target keywords] AND About Us NOT CONTAINS [“residential”, “consumer”, “B2C”, “staffing”]
Firmographic Layering
Stack semantic filtering with firmographic data for even more precision:
- Employee count: 50–500 (mid-market sweet spot)
- Revenue: $10M–$100M
- Geography: United States, Canada
- Funding stage: Series A–C (active growth mode)
Multi-Segment Filtering
For campaigns targeting multiple buyer personas:
(Industry IN [Set A] AND About Us CONTAINS [Keywords A]) OR (Industry IN [Set B] AND About Us CONTAINS [Keywords B])
Common Mistakes to Avoid
- Too few keywords (fewer than 10): You miss valid accounts using different terminology
- Too many keywords (more than 30): False positives from tangential mentions increase
- Skipping validation: Never deploy to full TAM without testing on 100 known accounts first
- Using your language: Build keywords from how customers describe themselves, not how you describe them
- Setting and forgetting: Review keyword lists quarterly — market language evolves
How This Fits Into Your GTM System
List building accuracy is the foundation — everything else depends on it. Once you have a 90–95% accurate list, the next step is combining it with signal detection and AI personalization to reach the right people at the right time. For the full picture of how list building connects to a complete outbound engine, read our guide to building a signal-led outbound system that generated $1.2M in pipeline with 90% automation.
Quick Reference: Filter Template
Use this structure for every new campaign:
- ICP Summary: One clear sentence describing your ideal customer
- Dream Accounts: 3–5 companies to model keywords from
- Industry Filter: 5–10 industry buckets (cast wide)
- About Us Keywords: 10–30 high-signal semantic terms
- Negative Terms: 3–5 exclusion terms to reduce false positives
- Firmographic Layer: Employee count, revenue, geography
- Validation Target: 85%+ accuracy on 100-company sample
Conclusion
The difference between a 50% accurate list and a 90%+ accurate list isn’t just better metrics — it’s the difference between a campaign that generates pipeline and one that burns your sender reputation. LinkedIn About Us semantic filtering is the most underutilized lever in B2B outbound. It costs nothing extra, takes 30–60 minutes to set up, and compounds across every campaign you run.
Build your keyword bank from your dream accounts, combine it with industry buckets, validate before deploying — then run it against your full TAM. Explore more UpSkillGTM frameworks to build the complete GTM system around your high-accuracy list.
Frequently Asked Questions
What is semantic filtering in B2B list building?
Semantic filtering uses the specific language companies use to describe themselves (from their LinkedIn About Us description) as a targeting mechanism. Instead of relying on broad industry codes, you filter for exact product terms, category language, and pain-point vocabulary that your ICP uses. This achieves 90–95% list accuracy vs. 50–60% with industry codes alone.
Do I need Clay to use LinkedIn About Us filtering?
Clay is the recommended tool because it has native LinkedIn company enrichment and makes the About Us field easy to filter on. However, you can also use Apollo exports to Google Sheets and apply CONTAINS formulas, or build a custom scraper. The methodology works with any tool that can pull and filter on text fields.
How many keywords should I include in my About Us filter?
Target 10–30 keywords. Fewer than 10 and you’ll miss valid accounts that use different terminology. More than 30 and you start getting false positives from tangential mentions. Build your keyword bank from 3–5 dream accounts, then validate on a 100-company sample before deploying to your full TAM.
How often should I update my keyword list?
Quarterly review is recommended. Update immediately if you notice high disqualification rates in campaign responses, significant market language shifts, or new competitor terminology entering your space.
