Pardot Scoring Model: How to Build One Sales Trusts

A Pardot scoring model is the set of rules that turns prospect behavior into a number (score) and prospect fit into a letter (grade), so Sales sees the right people at the right time. Build it in four moves: agree on what a qualified prospect looks like with Sales, give points to the actions that signal intent, grade on firmographic fit, and set thresholds that trigger assignment. Then audit it on a schedule, because a model nobody revisits drifts until reps stop believing it.

Pardot is now called Account Engagement, but most teams, and most searches, still use the old name. This guide uses both. It covers how scoring and grading work, how to set point values and thresholds, where models quietly break, and when machine-learning scoring is worth turning on.

TL;DR

• Score measures behavioral interest (a number). Grade measures fit with your ideal customer (a letter). A useful model uses both.
• Start from a shared definition of “qualified” with Sales, then set point values and thresholds. Salesforce’s own starting thresholds are 100 points and grade B-, to be adjusted for your business.
• Watch for double scoring, retroactive changes to baseline rules and missing score decay. Pardot does not decay scores out of the box.
• Einstein scoring can complement rule-based scoring, but it needs enough historical data to work.
• Review the model against closed-won and closed-lost data on a recurring schedule.

Table of Contents

• Score vs. grade: the two halves of the model
• Step 1: Define “qualified” with Sales before you touch a setting
• Step 2: Choose the signals and point values
• Step 3: Build the grading profile
• Step 4: Set thresholds that trigger action
• Where to configure it in Account Engagement
• Pitfalls that quietly break a Pardot scoring model
• Rule-based vs. Einstein scoring
• How to test and refine the model
• Perspective
• Need a scoring model built and maintained?
• FAQ
• Sources

Score vs. grade: the two halves of the model

Scoring and grading answer different questions. Salesforce describes the score as a numeric measure of behavioral interest, and the grade as a letter reflecting demographic or firmographic information. The practical reading: score tells you how interested a prospect appears, grade tells you how attractive that prospect is to your business.

That split matters because each one fails alone:

• High score, poor grade. A student, a competitor or a job seeker who reads every blog post and opens every email. Lots of activity, no revenue.
• High grade, low score. A VP at your ideal account who has barely touched your content. Right company, no signal yet, so a rep should not call yet but marketing should keep nurturing.
• High on both. The prospect worth a same-day follow-up.

How scoring works

In Account Engagement, scoring assigns numerical points to prospect activities to measure implicit interest, and the baseline scoring system is customizable. Baseline points attach to activities such as email opens, webinar attendance and form submissions. On top of that, automation rules can raise, lower or reset a score, and completion actions can adjust points when a prospect interacts with a specific asset.

How grading works

Grading starts from a different place. Prospects begin at grade D and move up or down according to profile criteria, on a scale from F to A+. The default grading profile includes company size, industry, location, job title and department, and each can be edited to describe your ideal prospect.

The mechanics in automation rules are worth knowing before you build. Each profile criterion needs its own rule. A criterion marked Matches raises the grade, Does Not Match lowers it, and Not Known leaves it unchanged. That last state is useful: missing data should not punish a prospect, it should simply not help.

Step 1: Define “qualified” with Sales before you touch a setting

The most common reason scoring models fail is not technical. Marketing picks numbers alone, Sales never agrees with the output, and the model gets ignored. Salesforce Trailhead frames model building as discovery with Sales and Marketing, data collection and integration, defining ideal-prospect thresholds, automating engagement metrics, and monitoring and iterating.

Run the discovery as a working session, not a survey. Bring a list of the last 20 to 30 closed-won deals and a handful of closed-lost ones, and answer these together:

1. Which job titles and departments were involved in the deals you won?
2. What company sizes, industries and regions close at a healthy rate?
3. Which actions did those buyers take before the first sales conversation: pricing page, demo form, webinar, repeat visits, a reply to a nurture email?
4. What does a rep need to see to feel a prospect is ready for outreach?
5. What happens to a prospect who is a great fit but not yet active?

Write the answers down as the model’s specification. The rules you build are just that document expressed in Pardot.

If your CRM and marketing data do not line up yet (duplicate records, unmapped fields, inconsistent job titles), fix that first. Grading on a field that is blank for half your database produces noise. Our guide to building a working single customer view covers the data foundation a scoring model depends on.

Step 2: Choose the signals and point values

A model should score behavior in proportion to how closely it predicts buying. Opening an email is cheap and easy to trigger, including by privacy features and security scanners. Submitting a demo request is a deliberate act. Your point values should reflect that gap.

Below is an illustrative starting framework. These values are examples of relative weighting, not Salesforce defaults, and you should replace them using your own closed-won analysis.

Two principles keep tables like this honest.

Weight by stage, not by volume. Ten blog views should not equal one demo request. If a prospect can reach your threshold purely by reading content, your model will hand Sales researchers instead of buyers. Salesforce Ben recommends adjusting points for specific marketing assets using automation and being purposeful about score changes, which is the right instinct: a case study and a generic newsletter should not carry the same value.

Use negatives. Scoring is not only additive. Automation rules can lower or reset a score, so use them for unsubscribes, bounced emails, student or personal email domains and long periods of inactivity.

Step 3: Build the grading profile

Grade should describe your ideal customer profile and nothing else. Start with the default criteria, then edit each to match the accounts you actually win.

For a B2B SaaS company, a profile might look like this:

• Company size: the employee band where your deals close and your product fits.
• Industry: the verticals where you have references.
• Job title and department: economic buyers and champions ranked above individual contributors.
• Location: regions you sell into and support.

Because each criterion needs its own automation rule, keep the profile short. Four to six criteria you trust beat fifteen you cannot maintain. And lean on the Not Known behavior: when a prospect has no company size on record, leave the grade where it is and use progressive profiling on forms to fill the gap.

Step 4: Set thresholds that trigger action

A model is only useful when crossing a line does something. Salesforce Trailhead gives 100 points and grade B- as starting qualification thresholds, while stressing that they must be adjusted for industry, business goals, sales needs and feedback. Its worked example later qualifies prospects at 150 points and B+. Treat those as a place to begin a conversation, not as the answer.

A practical way to calibrate:

1. Pull scores and grades for closed-won deals at the time of first sales touch.
2. Find the score and grade range where most wins sit. That is your candidate threshold.
3. Check how many current prospects would cross it each week. If Sales cannot absorb the volume, raise it. If almost nothing qualifies, lower it.
4. Pilot with one team or region for a few weeks before rolling out.

Then make the thresholds do work. Salesforce recommends using automation to calculate scores and update statuses, categorizing prospects as cold, warm or hot, and applying lead-assignment rules so qualified prospects reach the appropriate sales team. A simple three-tier approach works well:

The warm tier is where the two-dimensional view earns its keep. High fit with low activity needs different content than high activity with poor fit. For a deeper treatment of dual thresholds, see our write-up on dual score lead scoring models for B2B.

Where to configure it in Account Engagement

The documented path is Account Engagement Settings, then Automation Settings, then Scoring, then Edit Scoring Rules. Baseline activity values live there. After you change a rule, Account Engagement shows the number of prospects whose scores may be affected, so read that number before you save.

For everything beyond the baseline (asset-specific points, negative scoring, grade adjustments, score-triggered assignments), use automation rules and completion actions. Keep a naming convention for them, such as “Score | Pricing page | +15”, so anyone auditing the account can read the model from the rule list.

Pitfalls that quietly break a Pardot scoring model

Baseline changes are account-wide and retroactive

This is the one that catches experienced admins. Salesforce Ben warns that changing baseline scoring applies across the account and can be applied to historical activity, which can cause score-triggered automations to fire for additional prospects. Edit a baseline value on a Friday and Monday could bring a flood of reps’ alerts for contacts who were never newly active. Before changing a baseline rule, check the affected-prospect count, pause or review automations triggered by score and communicate the change to Sales.

Double scoring and cancellation

Baseline rules and automation rules can stack. A form submission might earn baseline points and also fire an automation rule that adds more, or one rule can quietly cancel another. Salesforce Ben advises auditing both mechanisms for double scoring or cancellation. A quick test: create a test prospect, perform one action at a time, and confirm the score change matches your spec.

No decay by default

A prospect who clicked six emails two years ago should not look the same as one who did so last week. Pardot does not include score decay out of the box, so you need to build automation for it. Common patterns are a recurring automation rule that subtracts points from prospects with no activity in a set window, or a rule that resets the score after a long dormant period. Pick windows that match your sales cycle: a 30-day cycle and a nine-month enterprise cycle need different decay.

Scoring one person when buying is a group decision

Individual scores can understate a deal where five people from one account each engage lightly. If your motion is account-based, review account-level reporting alongside prospect scores, and consider a rule that flags accounts with multiple engaged contacts.

Rule-based vs. Einstein scoring

Rule-based scoring is transparent: you know exactly why a prospect has 85 points. Its weakness is that you set the weights by judgment. Einstein scoring takes the opposite approach.

Einstein Behavior Scoring produces a 0-to-100 score, ranks prospects, surfaces positive and negative predictive factors and includes score decay, with recent activities typically scoring higher than older ones. Salesforce Engineering describes it as machine learning that learns engagement patterns and buying signals rather than relying on manually assigned points, and says the model weights proactive actions like clicks and form submissions above email opens, and recent activity above distant activity.

There are data requirements to check before you plan around it. Salesforce says it needs one year of engagement data for connected prospects and at least 20 prospects linked to opportunities associated with accounts, can take up to 48 hours to produce scores and typically refreshes every four hours.

How to think about the choice:

For many teams the strongest setup is both: keep a simple rule-based score and a grade for transparency and routing, and use the Einstein ranking as a second opinion on which prospects to work first.

How to test and refine the model

A scoring model is a hypothesis about what buyers do. Test it. Salesforce Trailhead recommends monitoring the lead qualification model, fine-tuning it and iterating on data. Salesforce Engineering’s account of earlier unsupervised behavior scoring names limits worth remembering: trouble incorporating new signals, limited validation, difficulty using customer feedback and limited explainability. Those are reasons to validate any model rather than treat it as permanent.

A review routine that works:

• Monthly: check how many prospects crossed the threshold, how many Sales accepted and how many it rejected, and why. Rejection reasons are the fastest feedback you have.
• Quarterly: compare score and grade at first touch for won and lost opportunities. If lost deals score as high as won ones, your weights are wrong.
• After any major change: a new product line, pricing change, new content program or new region should prompt a review of signals and grade criteria.
• Always: log what you changed and when, so a shift in conversion can be traced to a cause.

If you want a sharper read, attribution context helps too. Self-reported “how did you hear about us” answers can reveal intent that clicks never show, and we cover turning them into pipeline signals in our piece on self-reported attribution.

Perspective

Most scoring models we inherit are not wrong in their math. They are wrong because nobody owns them after launch. Marketing built the points, Sales never helped define the thresholds and the CRM integration quietly dropped half the fields grading depends on.

We treat scoring as a system with an owner, not a settings page. That means the person who designs the model also helps build the automation, the routing and the reporting, then stays in the loop to tune it against pipeline. When strategy and execution sit with the same people, the model changes when the business changes, rather than six months later.

If your Pardot or Account Engagement instance has grown into layers of rules no one can explain, the fix is rarely more points. It is a clear specification, clean data and a review rhythm Sales trusts.

Service

Need a scoring model built and maintained?

If your team is running Account Engagement alongside Salesforce or another CRM and wants scoring, grading, routing and reporting working as one system, that is what our revenue operations work covers. We embed with your marketing and revenue team, design the model with Sales, build the automation and keep tuning it against pipeline.

Engagements run on monthly capacity, starting with Core capacity at $3,500 to $5,999 per month. See the pricing page for the Growth and Scale tiers, or look at our full capabilities.

FAQ

What is a good lead score threshold in Pardot?

There is no universal number. Salesforce Trailhead suggests 100 points and grade B- as a starting point and stresses adjusting for your industry, goals and Sales feedback. Calibrate against the scores of your past closed-won deals at first sales touch, then pilot before rolling out.

What is the difference between Pardot score and grade?

Score is a number that reflects how engaged a prospect is, based on activities like clicks, form fills and webinar attendance. Grade is a letter, from F to A+, that reflects how closely the prospect matches your ideal profile, such as company size, industry and job title. Use both together to decide who Sales should call.

Does Pardot lower scores over time automatically?

Not with rule-based scoring. You need to build automation to decay or reset scores for inactive prospects. Einstein Behavior Scoring, by contrast, includes decay and weights recent activity more heavily.

Is it safe to change Pardot scoring rules on a live account?

Yes, with care. Baseline changes are account-wide and can apply retroactively, which may trigger score-based automations for more prospects than you expect. Check the affected-prospect count Account Engagement shows after an edit, review automations tied to score, and tell Sales before you publish.

Sources

• Lead Scoring: Enhance Your Marketing Strategy (Salesforce Trailhead)
• Lead Grading Essentials for Effective Marketing (Salesforce Trailhead)
• Lead Qualification Model Development Guide (Salesforce Trailhead)
• Lead Qualification Model with Account Engagement Insights (Salesforce Trailhead)
• Einstein Behavior and Lead Scoring Overview (Salesforce Trailhead)
• Use Account Engagement Automation Rules to Grade Prospects (Salesforce Help)
• Pardot Scoring vs Grading: Pinpoint Hot Prospects (Salesforce Ben)
• The Complete Guide to Pardot Score (Salesforce Ben)
• Machine Learning Driven Sales and Marketing with Einstein Behavior Scoring (Salesforce Engineering)

Recommended

• Dual Score Lead Scoring Models for B2B: 80/60 Thresholds
• Revenue operations
• B2B Landing Page Optimization: A Practical Playbook
• Build a Working Single Customer View in 7 Steps for Marketing & Data