How to build a client database automatically for B2B sales

“Just give us a client database” — that is how half of all sales-automation requests begin. But a database on its own does not sell. What sells is a precise list of companies that genuinely need your product, with the right contact and a clear reason for the first touch. Building such a list manually is theoretically possible: a rep spends 10–15 minutes per company — finds the website, hunts for an email, checks whether they are even a fit. Thirty companies a day, and that is his entire working day instead of calls and meetings. Automated lead collection takes the routine away: searching, enrichment, verification and pushing into the CRM are done by the system, while the human talks to people. Let’s walk through how such a system is built step by step, and where the legal boundaries run.

It all starts with the ICP, not with automated lead collection

Before you write a single line of code, you have to answer who exactly you are looking for. An ideal customer profile (ICP) is not “everyone who might need our services” but a set of concrete filters: industry (say, wholesale building materials), size (20 to 200 employees, or turnover above 30 million hryvnias), region, the presence of a particular process or technology. A good quality test for an ICP: if a search based on it returns 50,000 companies, the profile is badly written. A realistic number for most Ukrainian B2B niches is anywhere from a few hundred to a few thousand companies. That is your market, and it is entirely possible to work it properly.

The second layer of an ICP is a buying-readiness signal, a trigger. A company has posted a vacancy for “sales manager, 5 openings” — it is growing and will soon need a CRM. It has won a tender — money and workload have arrived. It has opened a new warehouse, redesigned its website, obtained a licence, appeared in the registry as a new player. Such signals turn a cold contact into a warm one: you don’t write “hello, we do automation,” you write “we saw you are hiring a sales team — here is what a CRM for a team that size usually looks like.” The difference in reply rates is several-fold. That is exactly why customer search should be built around events rather than around an alphabetical list of companies from a directory.

Open sources for a B2B client database

Sources for B2B lead generation in Ukraine exist, and most of them are open. Public registries and services such as YouControl or Opendatabot provide the company name, tax ID, industry code, status, director and sometimes financial statements — the basis for filtering by industry and size. Industry catalogues, association directories and exhibition attendee lists give narrow, targeted selections. Tender platforms show who is buying and for how much. Job boards are the best source of growth signals: a company that is hiring is almost always changing something in its processes. LinkedIn helps you understand the company structure and the roles inside it, while maps and aggregators are useful if your clients work in retail or services.

The practical approach is not to squeeze everything out of a single source but to stitch several together: take companies from the registry by industry code and region, apply a size filter, pull in their websites, check a sign of activity (the site is alive, there are fresh vacancies or new tenders), and drop anyone who is already a client or on a blacklist. Always respect each source’s terms of use: some services offer an official API or a subscription export, and it is almost always cheaper and safer to take the data through a legal channel than to invent workarounds. Deduplication is a separate, underrated job: the same company easily ends up in the database three times under different spellings of its name, and the rep calls the same director three times.

Contact enrichment: from the website to the decision maker

Next comes enrichment. From a company name you have to reach the person who makes the decision. The chain usually looks like this: name, website, corporate domain, general contacts, specific role. A generic mailbox is the worst channel: an administrator sits behind it and your letter rarely goes any further. So you look for the decision maker: in a small business that is the owner or director, in a mid-sized one it is a functional head — commercial director, operations director, head of IT. The sources are the “Team” page on the site, professional networks, signatures in public documents, talks at industry conferences. One accurate named contact is worth twenty info@ addresses.

Every contact you find has to be validated: check that the domain exists, that the mail server accepts letters, that the address is not a spam trap. A single validation pass protects your domain reputation: if 20% of your emails bounce, mail providers will quickly start dropping all your messages into spam — and even the warm clients who are waiting for you will suffer. And one more rule: enrichment is better done on demand rather than for the whole database at once. There is no point in hunting down decision makers for five thousand companies when only two hundred will actually be worked this quarter. Selection first, depth second — not the other way round.

Lead scoring and sales automation inside the CRM

Scoring is simple arithmetic that decides who gets called first. You assign points for ICP fit (industry +20, size +15, region +10), for triggers (a fresh vacancy +25, a won tender +20), for contact quality (a decision maker’s name and title +15, only a generic address 0) and for negatives (company being liquidated −50, already a client −100). Leads scoring above the threshold land in the CRM automatically, in the rep’s queue, tagged with the reason they are interesting right now. The rest wait in the database and may “grow” later, when a new signal appears. The rep opens the CRM and sees not a dump, but ten companies for today, each with a ready-made reason to talk.

Why is a thousand cold contacts worse than a hundred precise ones? Because a rep can physically work through 15–25 contacts a day properly. Hand him a thousand random companies and you are guaranteed template blasts, burnout and reputational damage: the more irrelevant emails, the more complaints, and the worse the deliverability of everything that follows. A hundred companies that truly fit the ICP, with the decision maker’s name and a personal opening line (“you have opened a second warehouse in Lviv — stock accounting usually becomes painful exactly at that step”), produce a reply rate several times higher. Sales automation should cut the time spent preparing a touch, not the quality of the touch itself: the system brings the fact and the contact, the human adds the meaning.

The legality of B2B lead generation and personal data

Here it is important to be precise. Data about a legal entity — name, registration code, address, type of activity, corporate phone and email on its own domain — is information about a company, and working with it is far simpler. But the moment you store a specific person’s first name, surname, job title and personal contact, that is already personal data, even if you found it in the public domain. Data protection rules (in Ukraine, the dedicated law; for EU clients, GDPR) require a lawful basis for processing, transparency and the ability to delete the data on request. In B2B people often rely on legitimate interest — but that only works when the offer is genuinely relevant to the person’s role, you are ready to explain where you got the contact, and every message contains an easy way to opt out of further communication. A mass mailing to private addresses with no relevance at all is spam, with all the consequences that entails.

A separate word about purchased databases. Offers along the lines of “a million contacts for twenty dollars” are almost always leaks: stolen or illegally copied data. Using them is dangerous for three reasons. First, the direct risk of liability for unlawful processing of personal data. Second, the destruction of your domain’s reputation under a wave of complaints, after which even letters to existing clients stop arriving. Third, the most mundane — quality: the contacts are dead, the job titles have changed, and the conversion rate tends toward zero. The honest route is slower but delivers a predictable result: a clear ICP, open sources, enrichment, scoring, personalisation. These are exactly the lead-collection systems we build at Devlly — parsing open sources, enriching contacts, scoring them and pushing them into the CRM automatically, so that a rep starts the day with a conversation rather than a search.