How to find prospects who are not in Apollo or ZoomInfo

Apollo has roughly 275M contacts. ZoomInfo has more. Sales Navigator has the largest B2B graph in the world. Yet for certain B2B ICPs, all three return mostly noise:

• Boutique retailers (independent fashion, home interior, jewellery)
• Hospitality operators (restaurants, cafes, hotels, bars)
• Local service businesses (dentists, salons, repair shops, agencies under 5 employees)
• Niche manufacturers in specific industries (specialty food, B2B components)
• Sole traders and family businesses in any vertical

These ICPs share traits: small teams, often no formal LinkedIn presence beyond a personal profile, no corporate email setup (they use Gmail or business-domain personal email), and they do not voluntarily submit their info to B2B databases.

If your client targets one of these ICPs and your sales team is "running outbound on Apollo lists", you are losing 70-80% of your TAM to a sourcing problem you did not know you had.

The Jamezz case study

Jamezz is a Dutch hospitality SaaS targeting restaurant operators in NL and Belgium. Restaurant operators do not show up cleanly in Apollo. Their company emails are often info@restaurant-name.nl that bounce. Their LinkedIn profiles, if they exist, often say "Owner" without a clear job-history trail.

Standard B2B outreach got Jamezz nowhere. So we built a custom sourcing pipeline.

Step 1: Google Maps scraping

Google Maps has nearly every restaurant in the NL with name, address, phone, website, and category. We scraped Google Maps results for "restaurant" within Dutch postal codes using Outscraper (you could also use SerpAPI or a custom Apify actor). Output: a CSV of ~12,000 restaurants with website URLs.

Step 2: Website parsing with Claygent

For each restaurant URL, we ran a Claygent prompt (Clay's AI agent) that visited the homepage and contact page and extracted:

• Owner or manager name (often in "About us" or "Contact" section)
• Email address (if listed publicly)
• Whether they list multiple locations (signal for ICP fit: chain vs single venue)
• Their current ordering or POS system (visible in checkout flows or footer logos)

This worked for ~70% of restaurants. The other 30% had no useful information on their websites.

Step 3: Email waterfall on the names we found

For restaurants where we got an owner/manager name but no email, we ran our standard Clay email waterfall (Findymail → Datagma → Apollo → Hunter verification). For sole-proprietor restaurants this hit at ~50%. For chain restaurants with proper corporate setups it hit at ~85%.

Step 4: ICP scoring before campaign

We scored each restaurant on 6 signals: location quality (city center vs suburb), social proof (review count and rating), existing tech stack visible on website, number of locations, cuisine type, and likely revenue band. We dropped the bottom 60% and kept ~3,400 high-fit restaurants for the campaign.

Step 5: Cold email with restaurant-specific personalisation

Each email referenced something concrete from the data we scraped:

• "Saw {{restaurant_name}} uses {{competitor_pos}}"
• "Your {{cuisine_type}} menu at {{location}}"
• "{{review_count}} reviews and a {{rating}} on Google"

Real signal, not template fill-in. Reply rate hit 11.4% across the campaign, well above the 3% benchmark Jamezz had seen from generic outreach.

Result

81 qualified leads from an unreachable ICP in 60 days. USD 190K+ pipeline. The full case study is here.

How to apply this to your own outbound

The general pattern works for any "unreachable" ICP:

1. Find a public directory or map service that catalogues your ICP. Google Maps for retailers/restaurants. Industry association membership lists. Niche directories (Houzz for home services, Tripadvisor for tourism, Yell for UK trades).
2. Scrape that directory at scale. Outscraper, Apify, custom Selenium. Cost: usually €50-200 for tens of thousands of records.
3. Visit each result's website with AI agents to extract data not in the directory.
4. Run your email waterfall on the names you collected, expecting a lower hit rate than corporate B2B.
5. Score before campaigning, accepting that you will drop a big share. The remaining list is gold.

This is more upfront work than a 5-minute Apollo search. It produces 5x the hit rate on the right ICP. If your TAM is in this long tail, the math always works in favour of custom sourcing.

What we will not do (and you should not either)

• Buying scraped lists from list resellers. Quality is uniformly bad and you are paying for data you could collect cleanly yourself in a day.
• Scraping LinkedIn directly. LinkedIn ToS, ban risk on your sourcing account. Use PhantomBuster for legitimate Sales Nav exports if needed, not direct LinkedIn scraping.
• Spamming role-based emails (info@, contact@). High bounce rate, often goes nowhere. Better to invest in finding the personal email of the actual decision maker.

The unreachable ICP problem is solvable. It just requires recognising that your sourcing layer is the bottleneck, not your copy or your sequencing. Most teams blame the wrong thing.