Pipedrive Lead Scoring With AI: What Is Built In and How to Build the Rest

    What Pipedrive scores out of the box, where that stops, and how to build AI scoring for leads that your reps will actually trust.
    GuideSander Vergouwen14 minUpdated: 16 November 2026

    In short

    Yes, Pipedrive has lead scoring on deals. On Premium and higher plans, its Scores feature gives each deal a score from 0 to 100 on up to ten criteria you set, with one active score per pipeline. An AI engagement score rates deals high, medium or low. For leads in the Leads Inbox, or for scoring free text with a reason, you connect an AI model through Make.com.

    Key takeaways

    • Pipedrive scores deals natively with Scores on Premium and higher. Leads in the Leads Inbox and free text are where you build your own
    • On deals, try native Scores first. It is the cheapest rule-based route and costs nothing to run
    • Eight signals are enough for a first model. Thirty signals mostly add noise
    • A score nobody acts on is just another field. Tie every score band to an automation
    • Clean data first. An empty company size field cannot be scored, however good the model is

    Source: Sales Surge Pipedrive implementations, Pipedrive features as of November 2026As of: 16 November 2026

    In this article
    1. Does Pipedrive have lead scoring?
    2. Rule-based or AI scoring: which one when?
    3. A worked scoring model with eight signals
    4. The prompt and the output schema
    5. Where the score shows up in Pipedrive
    6. How the connection with Make.com runs
    7. Building it step by step in three weeks
    8. How do you check whether the scoring is right?
    9. GDPR and explainability
    10. When AI lead scoring makes no sense
    11. Common mistakes
    12. How we build it
    13. Glossary
    14. Frequently asked questions

    Most B2B teams score leads on gut feeling, in a spreadsheet or with a points system someone set up two years ago. The question we get most is simple: does Pipedrive have lead scoring at all? This guide answers that first, then shows when plain rules are enough, and finally walks through a complete AI scoring model with the signals, the weights, the prompt, the output format and the place in Pipedrive where the score ends up.

    Does Pipedrive have lead scoring?

    Yes, on deals. At the time of writing (November 2026), Pipedrive has a feature called Scores on Premium and higher plans. A deal admin sets up to ten criteria on deal and activity fields and puts each one in a group: highly positive, slightly positive or negative. Pipedrive adds up the matches and shows a score from 0 to 100 on the deal. Only one score can be active per pipeline.

    Next to that, Pipedrive gives deals an AI engagement score: a high, medium or low likelihood of success, based on patterns in won and lost deals and refreshed daily. Pulse, the prospecting area where Scores lives, also ranks prospects by buying intent. Parts of Pulse were in beta at the time of writing. Win probability is something else again: a fixed percentage per stage for the weighted value in your forecast.

    So what is missing? A score on leads before they become a deal, a score on plans below Premium, and a score that reads free text such as a form message or call notes and writes a reason next to the number. That is where you build your own: with custom fields and rules, or with Make.com and an AI model. Plans and features change regularly, so check the current Pipedrive plan page.

    In PipedriveWhat it doesIs it lead scoring?
    Scores (Premium and higher)A score from 0 to 100 per deal on up to ten criteria you set, one active score per pipelineYes, on deals. Not on leads in the Leads Inbox
    AI engagement scoreHigh, medium or low likelihood per deal, refreshed dailyPartly: it ranks deals, but on criteria you do not set
    Pulse (in beta at the time of writing)Prospecting feed that ranks prospects by buying intentPartly, and availability differs per account
    Win probability per stageA fixed percentage per stage for the forecastNo, it is the same for every deal in a stage
    Leads InboxUnqualified leads with labels and sourceNo, and this is where a score of your own helps most
    LeadBooster (add-on)Chatbot, web forms, live chat, ProspectorNo, it brings leads in, it does not rank them
    Custom fields with Make.com and an AI modelYour own fields, rules and a model that reads free textYes, also on leads, with a reason per score

    Rule-based or AI scoring: which one when?

    Start with rules unless you have enough history. A rule-based score adds points for fixed criteria, for example 20 points for a decision maker. It is cheap and fully explainable. On deals, native Scores is the cheapest route if you are on Premium or higher. For leads, or on a lower plan, you build the same rules with custom fields and Pipedrive automation. AI scoring reads free text and combinations of signals, but needs examples to calibrate on and costs a little per call.

    Our rule of thumb: AI scoring starts to make sense from about 50 won and 50 lost deals with the key fields filled in, and roughly 100 new leads per quarter. Below that, the model has too little to learn from and a good rule set performs just as well.

    Rule-basedAI scoring
    Data neededA few filled fields per leadFields plus free text and history of won and lost deals
    Minimum historyNoneEnough won and lost deals to learn from, see the rule of thumb above
    MaintenanceAdjust the points by hand every quarterRecalibrate the prompt on new outcomes
    ExplainabilityFully: you see which rule added pointsOnly if you ask for a reason with every score
    Running costNone beyond Pipedrive, with native Scores on PremiumMake operations plus a small AI cost per lead
    Good atClear ICP fitForm messages, notes, mixed signals

    A worked scoring model with eight signals

    This is the model we start from with most clients. It scores leads, so it also works before a lead becomes a deal. It uses eight signals. Seven add points, together worth 100 at most. One subtracts points. We set the weights with the sales team based on recent won deals, so treat the numbers below as a starting point, not a standard.

    SignalExamplePointsSource in Pipedrive
    Job titleDecision maker or direct influencer20Person field, filled by form or enrichment
    Company sizeWithin your ICP band15Organisation field, enrichment via Apollo.io
    IndustryOne of your core industries10Organisation field
    Lead sourceInbound request instead of a cold list15Lead source field
    Website visitPricing or case page in the last 14 days10Visitor data via webhook
    Email replyReplied to a sales email15Email sync
    Meeting bookedA call or demo is planned15Activity of type meeting
    Negative signalFree email address, student, competitorminus 20Person and organisation fields

    The prompt and the output schema

    The model gets the signal values, your ICP in a few sentences, the weights above as guidance and five short examples of won and lost leads. We keep the temperature low, so the same lead gets the same score tomorrow.

    Prompt

    You score B2B leads for the sales team.
    Use only the data below. If a field is empty, say so in the reason.
    Start from the points table. You may adjust by at most 15 points
    based on the message or notes, and you explain why.
    
    ICP: {{icp_description}}
    Points table: {{weights}}
    Examples of won and lost leads: {{examples}}
    Lead: {{lead_fields}}
    Message or notes: {{free_text}}
    
    Return only JSON that matches the schema.

    Output schema

    {
      "score": 78,
      "reason": "Decision maker in core industry, visited pricing twice, plans migration this quarter",
      "confidence": "high",
      "signals_used": ["job_title", "company_size", "industry", "source", "website_visit"],
      "missing_fields": []
    }

    Make checks the answer before writing it back: the score must be a whole number between 0 and 100, confidence must be low, medium or high, and the reason may not be empty. An answer that fails the check goes to an error route and the lead keeps its old score.

    Where the score shows up in Pipedrive

    The score is only useful if it changes what someone does. We write it to three custom fields: AI score (number), AI reason (text) and AI confidence (single option). Native Pipedrive automation then reacts to the number.

    • 75 or higher: label Hot, a call activity for the owner due today and a Slack message
    • 50 to 74: into the nurture sequence, for example through Instantly.ai, and a check after 30 days
    • 25 to 49: stays in the Leads Inbox and is rescored when something happens
    • Below 25: archived with a reason, so you can find it again later

    We rescore on every real interaction (a reply, a booked meeting, a viewed proposal) and once a night for leads that changed. Scoring only at creation means the score is wrong a week later.

    To add follow-up drafts and call summaries to Pipedrive as well, read Pipedrive AI automation.

    How the connection with Make.com runs

    1. 1. Collect

      Custom fields hold firmographics and source. Activities and email sync hold behaviour. Visitor data comes in through a webhook.
    2. 2. Trigger

      A Make scenario starts on a new lead, a changed field, a logged activity or the nightly run.
    3. 3. Enrich and clean

      Apollo.io fills missing company size or industry. Make normalises values, so 11 to 50 and 11-50 are the same band.
    4. 4. Score

      Make calls the model with the prompt above and checks the JSON.
    5. 5. Write back

      The score, reason and confidence land in the custom fields. Pipedrive automation takes it from there.

    How we set up and maintain Make scenarios is explained on our Make.com page.

    Building it step by step in three weeks

    Week 1: data and model

    • Review existing fields and how often they are filled
    • Choose the eight signals with your best reps
    • Create the custom fields and the points table
    • Turn on enrichment for missing firmographics

    Week 2: Make.com and the prompt

    • Build the scenarios on the Pipedrive triggers
    • Write the prompt with your ICP and examples
    • Test on 40 historical leads, 20 won and 20 lost
    • Build the error route and the JSON check

    Week 3: live and feedback

    • Switch on scoring for all new leads
    • Explain to the team what a score means and what to do with it
    • Add a field where reps mark a score as right or wrong
    • Set up a weekly view of the checks below

    How do you check whether the scoring is right?

    You check it against what happened afterwards, not against what feels right. We look at four numbers. The first one is the most important, and the target is our own starting point, not an industry benchmark.

    CheckDefinitionStarting target
    Precision of high scoresShare of leads scored 75 or higher that reach a qualified stage or a booked meeting within 30 daysAbove 75% after the first 30 days
    CoverageShare of deals won this quarter that had a score of 75 or higher before the first meetingRising every quarter
    Speed on hot leadsTime between the lead being created and first contact, for scores of 75 or higherSame working day
    DisagreementShare of scores that reps mark as wrongFalling after each recalibration

    Is precision low, then the model scores too generously: raise the threshold or tighten the prompt. Is coverage low, then you miss good leads: check which signal those deals had that the model ignored.

    GDPR and explainability

    Lead scoring in B2B usually rests on legitimate interest, but that does not make it free of rules. Mention in your privacy statement that you score leads and on which kind of data. Use only business data that is relevant to the sale, never special category data or private social media.

    Keep a person in the decision. A score decides who gets called first, not who is excluded. Store the reason with every score, so you can explain it when someone asks to see their data. Choose an AI provider with a data processing agreement and check whether inputs are used for training. This is not legal advice; check your own situation with your privacy officer.

    When AI lead scoring makes no sense

    We regularly advise against AI scoring. These are the cases:

    • Fewer than about 100 new leads per quarter. Your reps can read every lead, so a score adds little.
    • Too little history to learn from. Use the points table as rules instead, or native Scores on deals.
    • Key fields are mostly empty. Fix the intake first, then score.
    • Nobody acts on the score. Without an automation or a routine attached, it is just another field.

    Common mistakes

    Starting with AI before the data is right

    If a third of your company size field is empty, no model can say anything sensible. Clean data beats a clever model.

    Too many signals

    Start with eight. Add a signal only when you can show which deals it would have scored differently.

    Leaving sales out

    Your best reps know which signals matter. Build the points table with them, or they will not trust the result.

    No feedback field

    Without a way to mark a score as wrong, you cannot recalibrate. One single option field is enough.

    How we build it

    Lead scoring is part of our Marketing Automation service, together with nurture sequences: from €2,200 with a lead time of 2 to 3 weeks. We start with your data, because that is where most scoring projects fail.

    Scoring is only as good as what comes in. Surge Forms writes form leads straight into Pipedrive with the right fields, and Surge Chat qualifies website visitors and sends them to Pipedrive. That gives the model clean input from day one.

    We have delivered more than 150 CRM implementations, and since July 2026 we are among the top 8 Pipedrive partners worldwide as a Platinum Partner. So we know where native Scores is enough and where it is not.

    For more on nurture, scoring and segmentation, see our marketing automation guide. How scoring fits a broader AI approach is covered in the AI-first CRM strategy.

    Glossary

    Lead score
    A number, here 0 to 100, that ranks leads by how likely they are to become a customer.
    Scores
    Pipedrive's own score from 0 to 100 on deals, on criteria a deal admin sets. Available on Premium and higher plans, one active score per pipeline.
    Win probability
    A percentage per stage in Pipedrive, used for the weighted forecast. It says nothing about a single deal or lead.
    Precision
    Of all leads with a high score, the share that turned out to be good.
    Coverage
    Of all good leads, the share that had a high score in time.
    Calibration
    Adjusting weights or the prompt so scores match what happened with real leads.

    Frequently asked questions

    Does Pipedrive have lead scoring?

    Yes, on deals. At the time of writing, the Scores feature on Premium and higher plans gives each deal a score from 0 to 100 on up to ten criteria a deal admin sets, with one active score per pipeline. An AI engagement score also rates deals high, medium or low. Leads in the Leads Inbox get no score, and neither does free text. For those you build your own with custom fields, or with Make.com and an AI model.

    What if I am just under 50 won and 50 lost deals?

    Start with the points table as plain rules and log the outcome of every lead from day one. Then the history you need builds up while the rules already do their job. When you cross the threshold, test the AI model on those logged leads first and compare it with the rules. Switch only if the model ranks the won deals clearly higher. Often the rules keep doing fine for deals and AI only adds value on free text.

    What does it cost to have lead scoring built in Pipedrive?

    Lead scoring falls under our Marketing Automation service, together with nurture sequences, from €2,200 with a lead time of 2 to 3 weeks. Running costs are small: Make.com operations per scored lead and a small amount per call to the AI model. A rule-based version without AI has no running cost beyond your Pipedrive plan. Are you on Premium, then try native Scores on your deals first.

    Is automated lead scoring allowed under the GDPR?

    In B2B it usually rests on legitimate interest, as long as you mention it in your privacy statement, use only business data that is relevant to the sale and keep a person in the decision. A score should decide who gets called first, not who is excluded. Be careful with free text: a form message can contain more personal data than you need. Check your own situation with your privacy officer.

    Which AI model do you use for scoring?

    Any current model that reliably returns structured JSON will do. We choose per client, based on where the data may be processed and on cost per call. More important than the model are the prompt, the examples of won and lost leads, a low temperature so scores stay stable, and a check in Make that rejects any answer outside the schema.

    How often is the score recalculated?

    On every real interaction, such as an email reply, a booked meeting or a viewed proposal, and once a night for leads whose fields changed. Scoring only when a lead is created means the score is outdated within a week. Rescoring only changed leads keeps the number of Make operations and AI calls under control.

    What if reps do not trust the score?

    Then the score is useless, so plan for it. Show the reason next to every score, build the points table together with your best reps and add a field where they mark a score as right or wrong. Review those marks every month and recalibrate. Trust grows when reps see that their feedback changes the model.

    Citations & Sources

    The following sources were referenced in creating this article. Links marked with require confirmation before publication.

    Official partners

    Click a badge to sign up with our partners:

    Sander Vergouwen

    About the author

    Sander Vergouwen

    Co-founder of Sales Surge

    LinkedIn

    Want lead scoring in your own Pipedrive?

    Book a call. We look at your pipeline data and tell you whether rules or AI fit your volume, and what building it costs.