78 million domains: building a B2B lead database that works

It’s Monday morning and the sales team is staring at a spreadsheet the company bought last spring. Twelve thousand rows. Half the phone numbers ring out. A third of the websites now show a “this domain is for sale” page. The “industry” column says “Retail” for almost everyone. The reps grind through it anyway, and by Friday they’ve had four real conversations and three people asking how they got this number.

The reps aren’t the problem. A purchased list of unknown age and origin isn’t a B2B lead database. It’s a lottery ticket. This article explains how we build a B2B lead database from the open web, why only a fraction of 78 million domains ever reaches a salesperson, and how every company on the list gets a pitch written for it instead of a mass email.

Where 78 million domains come from

The raw material for the SCRAPUS company database is Common Crawl, a public archive that has been crawling the web for years. From it we pulled roughly 78 million domains.

Let’s be clear about what that number means. 78 million domains is not 78 million prospects, or even 78 million companies. It’s a queue. It includes abandoned sites, personal blogs, parked domains, placeholder pages, and spam networks built to game search engines. Crawling all of it at normal speed would take about a year. So we work through the queue in order, starting with the regions and top-level domains that a given customer actually sells into.

Here’s what one pass looked like. We checked about 300,000 websites. Around 214,000 of them, roughly 71%, were live businesses. For every live site we identified the country and the industry. The rest were dead domains, placeholders, junk, or sites hidden behind bot protection.

That ratio is the whole lesson. If your team works from raw domain lists, nearly a third of its effort is wasted before anyone picks up the phone.

What SCRAPUS learns from a company’s website

SCRAPUS reads each live site the way a careful intern would, just faster and without getting bored. Data collection doesn’t call a large AI model, so it costs nothing in AI usage.

What we collect	Why sales cares
Email addresses and phone numbers	The first point of contact
Phone type: mobile or landline	Whether messaging is even an option
Legal entity and registration details	Separates a real company from a storefront
Country	So outreach follows local law and business hours
Industry	So you can segment and say something relevant
Social profiles and team pages	Live channels and the people who make decisions

Industry isn’t guessed by a large model for every site. A lightweight classifier, trained on labeled examples, does the job. That’s why tagging hundreds of thousands of sites costs pennies rather than thousands of dollars.

Telling a real business from noise

A website is an unreliable witness. SCRAPUS separately detects:

• Dead sites: no DNS record, no response, or an error on the home page.
• Placeholders: “coming soon,” “domain for sale,” “registration expired.”
• Doorway pages and spam networks built only for search engines.
• Someone else’s site behind a redirect. If a domain forwards to a registrar or a directory, those contacts are not attributed to the company.
• Sites behind bot protection. Live, but hidden behind a browser check.

That last category matters. Most scrapers would mark these as dead. When we checked samples by hand, many of the “dead” sites turned out to be working businesses behind a protection screen.

Crawling in stages

To avoid losing live companies without burning money, the crawl runs in stages:

1. Fast pass. Picks up everything that opens easily and tags problem sites by type of failure.
2. Patient retry. Slower, one request at a time, a few hours later, for sites that timed out or were overloaded. A share of them come back.
3. Real browser. For sites that build pages with scripts or sit behind a check.

We also follow a courtesy rule. Many small sites share hosting, with hundreds of businesses on a single server. Hit them all at once and the host will rightly block you. SCRAPUS caps concurrent requests per server, not just per site, and spaces them out.

Turning a lead database into sales, not spam

Building the database is half the job. The other half is where most outbound programs fail: the first touch.

Cold data stays out of your CRM

We keep cold data outside the CRM. Only live leads move into the CRMUS CRM: people who replied, showed interest, or were hand-picked by a rep. In SCRAPUS you filter companies, click “Send to CRM,” pick a pipeline and a stage, and they land with your team.

Why bother? Dump ten thousand cold companies into a CRM and the pipeline becomes a graveyard, the reports stop meaning anything, and reps stop trusting the system. When the CRM holds only real conversations, a manager sees what’s actually happening.

A tailored pitch, not a blast

Each selected company gets a message built from what its own website says: what it does, what it sells, where it operates. Here’s a hypothetical example:

A company sells industrial equipment and takes orders through a web form. The first email talks about exactly that, how to answer inbound requests faster and stop losing them between email and chat, instead of listing every feature we have.

The AI sales agent in SALEUS drafts the message, but it doesn’t invent one from scratch every time. It works from a template your sales lead approved in advance and fills in facts about the company. Early on, anything involving pricing, meeting times, or a formal proposal goes out only after a human says yes.

The outreach rules are strict:

• Messages go out only during the recipient’s local business hours.
• A daily sending cap.
• A suppression list: anyone who asks us to stop never hears from us again.
• A reply from the prospect stops the automated sequence immediately; from there it’s a live conversation.
• Disqualifying a lead requires human confirmation.
• The AI introduces itself as an assistant and never pretends to be a person. Some jurisdictions now require that, and it’s the honest thing to do anyway.

Is cold email legal? What to keep in mind

This isn’t legal advice. Before you launch in a new country, have a local lawyer read your email. That said, a few habits lower the risk almost everywhere, and we follow them by default.

Four rules for a respectful cold email:

1. Write to a company’s general address (info@, sales@, office@), not an employee’s personal inbox.
2. One email plus one or two follow-ups at most. Not a drip campaign that never ends.
3. Say clearly who you are: company name, address, phone number.
4. Include a simple unsubscribe that actually works, and never contact people who used it.

The rules vary a lot by country:

Market	Business-to-business email, in short
United States	Allowed without consent under CAN-SPAM, with an opt-out and a physical mailing address
United Kingdom	Allowed to corporate addresses with an opt-out; not to sole traders or partnerships
European Union	Stricter in many member states; Poland, for example, requires consent even for business addresses

Our rule of thumb follows from this. We don’t cold-text and we don’t cold-call a scraped list. The first touch is a careful email or a message to a company’s public business channel. A phone call comes after someone has replied and agreed to talk.

Step by step: launching your own lead database

1. Define your ideal customer. Country, industry, size, and signals on the website (an online store, a request form, a careers page).
2. Crawl the regions you sell into. SCRAPUS checks the sites, drops dead ones and junk, and tags country and industry.
3. Build a segment. Filter by industry, country, and whether there’s a mobile number or an email.
4. Approve one first-touch template per segment, with company facts filled in automatically. Your sales lead reads it and signs off.
5. Set the outreach rules. Business hours, a daily cap, a suppression list, one or two follow-ups.
6. Turn on the sequence. Replies arrive in one shared inbox, and live leads flow into the CRM.
7. Watch what works. Which segments reply, which openers start a conversation. Then the calls begin, and it helps when every call turns into a summary, a next step, and a task.

Pre-launch checklist

• Writing to general company addresses, not personal ones.
• Sender’s company name and contact details are in every email.
• Unsubscribe link tested and working.
• Suppression list connected.
• Template approved by the sales lead.
• Business hours and daily cap configured.
• Local rules for each target country checked.
• Only people who replied are sent to the CRM.

Other places to find buyers

Websites aren’t the only public source. In logistics, construction, and wholesale, many companies post what they need in industry group chats. We covered how to find those requests without spam or bought accounts in Telegram lead generation without spam.

The bottom line

78 million domains makes a good headline, but the value isn’t in the number. It’s in a system that pulls live businesses out of a huge queue, sets the junk aside honestly, collects contacts and industry at no AI cost, and then writes to each company about something that matters to it, politely, during business hours, with an unsubscribe that works. Your CRM ends up holding only people who actually want to talk.

Want to see what the company database looks like for your industry, and what one of those companies would receive from you? Book a demo and we’ll walk through your segment. If you’re weighing the cost, the pricing page explains how plans are structured.