An illustrative example: the sales manager of an IT services company in Vilnius leads a team of three salespeople. Each month they get more than 200 leads from the website, LinkedIn and partner referrals. In theory, a great problem to have. In practice, a disaster. The three salespeople try to reach every lead, but many calls go to contacts who are "just looking", while a director who is actually ready to buy waits four days for a reply until another vendor gets to him first. Over a quarter, the team can lose more than one deal precisely because of slow follow-up, not because of price or quality.
The answer is not hiring a fourth salesperson. The answer is a lead scoring system that tells you automatically: "this lead wants to buy within two weeks, and that one is only gathering information." In this article I cover three levels of lead scoring (from Excel to predictive AI), six scoring dimensions adapted to the Lithuanian B2B market with specific weights, a comparison of tools with 2026 prices, and how to build a custom AI scoring model on your historical sales data in two weeks.
What is lead scoring, and why B2B companies in Lithuania still skip it
Lead scoring rates how "hot" each lead is against several criteria at once: firmographics (what the company is), behavior (what they do on your website and in email), the decision-maker's job title and intent signals. The result is a score from 0 to 100 that shows how likely the lead is to become a customer in the next 30–60 days.
In the Lithuanian B2B market, lead scoring is still not the norm for a few reasons: many companies with 5–30 employees think it is for large corporations with tens of thousands of leads; others know about scoring but consider it too expensive or too complex to set up. In 2026 neither assumption holds: a custom AI scoring model can be built in two weeks and run on roughly €30–60/month in platform costs (for example, n8n Cloud Starter is €20/month billed annually, as of October 2026, plus OpenAI API usage) if you use n8n and OpenAI instead of enterprise CRM licenses.
Manual vs rule-based vs AI scoring
Manual: a salesperson rates the lead by gut feeling after the first contact. This rating is inaccurate, inconsistent and suffers from the salesperson's "good day" effect. Often the manual rating of the same lead differs a lot between salespeople, or even on different days.
Rule-based: predefined criteria, for example +20 points if the company has 50–200 employees, +15 if they visited the pricing page. This method is more accurate than manual scoring but rigid: the rules do not change based on actual sales results and do not account for complex combinations of signals.
Predictive AI: a model trained on your historical sales data (500–2,000 deals) recognizes complex combinations a person would not notice: for example, that ICT companies with 20–60 employees that visited the page twice a week apart and downloaded a PDF convert several times better than those that visited only once. This method needs enough data but produces the fewest false positives.
3 levels of lead scoring: from Excel to predictive AI
Level 1: Excel with three columns (to get started)
Basic scoring can start with a simple three-column Excel sheet: company size (0–10), decision-maker's title (0–10), intent signal (0–10). Each lead gets a total of up to 30 points. Leads above 20 points are priority. This method needs no extra tools and lets you start prioritizing within an hour. The biggest drawbacks: manual data entry and criteria that quickly go stale.
Level 2: rule-based scoring in your CRM
HubSpot Professional and Pipedrive (with add-on tools) let you build automatic rule-based scoring: when a lead visits the pricing page, +20 points in the CRM automatically; when they download a PDF, +15; when they reply to the first email, +25. The system runs automatically, but the rules are created by hand and need periodic review. A good starting point for companies with 100–300 leads a month.
Level 3: a predictive AI model on historical data
Predictive AI scoring relies on your historical data. The model analyzes which leads became customers over the past 12–24 months and which did not, and learns from that difference to spot the same traits in new leads. With a history of 500+ deals, the model usually predicts more reliably than manual rating; it is worth measuring accuracy against your own historical data. More on building this level in the "Custom AI model with n8n" section below.
Which lead scoring signals matter most in Lithuanian B2B
Not every signal has the same value. Based on the specifics of the Lithuanian B2B market (dominant sectors: IT, logistics, manufacturing, professional services, small and mid-sized companies), the signals that best predict conversion fall into these four groups:
Firmographics from the Register of Legal Entities
Company size (number of employees from public Sodra (the Lithuanian social insurance fund) data, conveniently available through private directories such as Rekvizitai.lt), NACE activity code, company age and region are the four main firmographic variables for the Lithuanian market. The ideal customer for many IT service providers: a private limited company (UAB) with 15–80 employees, NACE code 62.xx or 70.xx, operating for 3–15 years, in the Vilnius, Kaunas or Klaipėda region. Companies that match this profile usually convert better than those that do not, and the scoring weights should reflect that.
Behavioral signals on the website and in email
Website pages and time: pricing page (+25 points), service description (+15), demo request (+40), email open (+10), PDF download (+20), return visit within 7 days (+15). Behavioral signals show the stage of the buying decision: a lead who visited three times in a week is actively comparing offers. A lead who opened the website three months ago and has not come back has probably already decided against you.
The decision-maker's title and position
Director, founder or CEO: +30 points. IT manager or operations manager: +20. Project manager: +10. Accountant or administrator: +5 (usually just gathering information for a manager). The title correlates directly with decision-making power, and in Lithuanian SMBs decisions are often made by one person, not a committee.
Engagement speed and activity
Leads who reply within a few hours usually convert much better than those who reply after a few days. A LinkedIn connection or follow: +10 points. Attending a webinar or event: +20. Replying to a reactivation email after 30 days: +15. These signals show active interest rather than passive curiosity and are the strongest short-term conversion indicators.
A practical lead scoring model: 6 dimensions with specific Lithuanian B2B weights
This model suits B2B service and SaaS companies in Lithuania. The total maximum score is 100 points. Internal grading: 70–100 is "hot" (call today), 40–69 is "warm" (nurture by email), 0–39 is "cold" (monitor).
1. Firmographics: up to 25 points
company size 15–80 employees (+15), NACE code matches your ICP (+7), company age 3–15 years (+3). The data is pulled automatically through the Rekvizitai.lt API for every new lead in the CRM.
2. Decision-maker's title: up to 20 points
Director/founder (+20), IT/operations manager (+15), project manager (+10), other (+5). Checked via LinkedIn or the job title field in the contact form.
3. Behavioral signals on the website: up to 20 points
Demo request (+20), pricing page (+12), return visit within 7 days (+10), PDF download (+8), services page (+5). Tracked via the Google Analytics or Plausible API.
4. Email engagement: up to 15 points
Replied within 2 hours (+15), opened the email three times (+8), clicked a link (+5). Data from your email tool via a Lemlist or Brevo API webhook.
5. Technology intent signals: up to 10 points
Uses your target technologies (for example, Pipedrive or HubSpot, a sign they are actively looking for automation solutions, +10), uses outdated systems (+5). Checked via BuiltWith or Apollo.io data.
6. Region and reachability: up to 10 points
Vilnius/Kaunas/Klaipėda (+10), other Lithuanian cities (+7), foreign company with operations in Lithuania (+5). Region directly affects the implementation cost and service options, so it is reflected in the score.
Lead scoring tools in 2026: prices and how to choose
Lead scoring can be set up in several ways. The choice depends on your lead volume, the CRM you already use and your budget:
HubSpot Professional / Enterprise: from ~€100/user/month, prices change
The Professional plan has rule-based lead scoring, while AI (predictive) scoring is on the Enterprise plan (as of October 2026). The upside: everything on one platform, with CRM, email, analytics and scoring. The downside: it is expensive (often several hundred € a month for a team, plus a one-off onboarding fee), and the scoring cannot be adjusted to Lithuania-specific databases. Suits companies that already use HubSpot and want a quick, integrated solution.
Pipedrive Premium / Ultimate: from ~€49/user/month, prices change
Pipedrive's higher plans include LeadBooster lead capture and the option to score leads by set criteria. However, the scoring is based on generic behavioral signals and does not integrate with Lithuanian Register of Legal Entities data. It works as a first stage, but optimal accuracy comes from adding a custom n8n integration.
n8n + OpenAI custom model: €30–60/month in platform costs
The best choice for B2B companies that want accurate scoring with Lithuanian data. n8n integrates with Rekvizitai.lt, website analytics, your email tool and CRM, and for every new lead it sends the data to OpenAI, which returns a score and a recommendation for the salesperson. The initial setup takes more work, but the monthly cost is much lower than CRM enterprise plans. More on CRM automation for small businesses with n8n integration.
How to build a custom AI lead scoring model with OpenAI and n8n in 2 weeks
Building a custom AI lead scoring model takes four components: historical data, a scoring prompt structure, an n8n workflow and a validation loop. Here is what it looks like in practice:
Week 1: preparing and analyzing the data
Export the last 500–1,000 deals from your CRM, both won and lost. For each deal, record: company size, sector, decision-maker's title, lead source (website, LinkedIn, partner), email reply speed, and whether there was a demo meeting. This data becomes the training dataset: from it the AI learns what separates won deals from lost ones and which signals were the early indicators.
Prompt structure for scoring
The OpenAI scoring prompt must include: context on your company's ideal customer profile (ICP), descriptions of the scoring dimensions with weights, the new lead's data and an instruction to return JSON with the score (0–100), the main reason and a recommended action ("call today", "send a nurture email", "monitor"). The prompt has to be specific to your business: a generic "rate this lead" prompt will give vague, inaccurate results that ignore the specifics of the Lithuanian market.
Week 2: the n8n workflow and validation
The n8n workflow is triggered when a new lead enters the CRM. The sequence: it collects firmographics from Rekvizitai.lt, pulls website analytics data (Google Analytics or Plausible API), checks email replies (Lemlist or Brevo), sends all the data to OpenAI, and writes the scoring result back to the lead field in the CRM. The salesperson sees the score, the main reasons and the recommended action in the CRM, with no manual work. In the first week, I recommend checking 20–30 scoring results by hand before letting it run unsupervised.
- 1A new lead enters the CRMn8n
- 2Firmographics are collectedRekvizitai.lt
- 3Behavioral signals are addedGoogle Analytics / Plausible · Lemlist / Brevo
- 4The model scores it from 0 to 100OpenAI
- 5Score, reasons and action on the CRM recordIn the first week, check 20–30 results by hand.
What results to expect after 60 days
Illustrative figures for a hypothetical scenario of a 3-person B2B sales team rolling out AI lead scoring with an n8n + OpenAI stack (not measured client results and not a guarantee; your numbers will depend on lead volume and process):
The key point: salespeople stop calling "cold" leads not because they ignore them, but because the system moves them into an automatic nurture email sequence, and the salesperson only sees those already showing active buying signals. Every call becomes more meaningful, because the salesperson knows this lead is worth talking to today.
Before scoringAfter 60 days
Leads called by salespeople
Lead sorting per week
Top 30% lead conversion
Illustrative figures from the table above, not measured client results. Where a range is given (60–70 leads, 4–5 hours, 18–25%), the more conservative value is shown.
Common mistakes when setting up lead scoring
Too many dimensions from day one
20–30 scoring variables at once create a system that neither salespeople nor managers understand. Start with 5–6 dimensions and add new variables only when you can clearly see the existing ones fail to predict conversion. Fewer variables mean clearer interpretation and more trust in the system.
The scoring is not updated regularly
A lead scoring model is not static. The market changes, your ICP changes, conditions in your sectors change. I recommend reviewing the scoring weights once a quarter: check whether the top 30% of leads are still the ones converting best and whether there are new signals the model does not yet account for. A model that is not updated regularly becomes less accurate within a year.
False positives are ignored
If the system keeps marking leads as "hot" that do not buy, you have a false positive problem. Every false positive wastes a salesperson's time and erodes trust in the system. Track the false positive rate (benchmark: below 20%) and adjust the weights if it is higher. In a well-maintained system this rate should fall over a few months.
Scoring is not aligned with marketing
Lead scoring works best when the marketing team knows which leads get a high score and creates content and campaigns that attract exactly that ICP. If marketing attracts one audience and the scoring consistently rates it as "cold", that is a conflict between system and strategy, not a technology error. It is solved with monthly sales + marketing reviews of the scoring statistics.
Privacy and GDPR: which data you can use for lead scoring
Lead scoring in Lithuania must comply with the GDPR. The most important points:
Using public data: data about legal entities (name, NACE code, number of employees) is public and can be used for scoring without consent. Note: a manager's first and last name is personal data; even if public, the GDPR applies, so you need a legal basis (usually legitimate interest) and must inform the person under Article 14 of the GDPR. Rekvizitai.lt is a private directory that aggregates public data from the Register of Legal Entities and Sodra, not an official register.
Behavioral data: for website analytics data and email opens, your privacy policy must state that such data is collected and used for sales purposes. For inbound leads (who filled in a form themselves), legitimate interest provides a basis for the first contact and scoring without separate consent.
LinkedIn data restrictions: automated mass scraping of LinkedIn is prohibited under LinkedIn's terms of use. You may only use publicly visible profile data (title, company) if the contact connected through a LinkedIn OAuth integration themselves. Scraping purchased databases is a separate GDPR risk that requires legal advice.
Anonymization in the n8n workflow: I recommend separating personal data (name, email) from the scoring components, so that only an anonymized company profile is sent to OpenAI, not individual personal data. This reduces GDPR risk and removes potential bias issues in automated scoring.
Aigentas service: predictive lead scoring from €1,500 setup
Aigentas.lt sets up B2B lead scoring systems for Lithuanian companies using n8n, OpenAI and Rekvizitai.lt integrations. I do not set up "AI scoring in general"; I set up specific scoring models with measurable metrics and a clear ROI calculation before you invest.
Free initial discussion. Write to me how many leads you get per month, which CRM you use and how you now decide whom to call first. I will read it and reply by email on whether AI scoring makes sense for you right now and what type would fit your volume. If needed, we will arrange a call.
Basic scoring (from €450). Rule-based CRM scoring with 5–6 dimensions, Rekvizitai.lt integration and an n8n workflow. The best choice for companies with 50–200 leads a month that want a quick, clear result. Setup in two weeks. More on business process automation with n8n.
Predictive AI scoring (from €1,500). A custom OpenAI model trained on your historical sales data, an n8n workflow with all integrations (Rekvizitai.lt, website analytics, CRM, email), a first-month validation stage and a quarterly model update. The best choice for companies with 200+ leads a month and 12+ months of sales history.