Account Scoring Best Practices for B2B | Abmatic AI

Introduction

Account scoring is the foundation of ABM prioritization. A good model surfaces the accounts most likely to buy, lets you focus sales resources, and evolves as you learn which signals matter most.

Unlike lead scoring (individual-level), account scoring weighs both firmographic fit and behavioral intent. This guide covers model design, implementation, testing, and refinement.

---

Two Dimensions of Account Scoring

Fit Score (Firmographic - Static)

How well does the account match your ideal customer profile?

Fit attributes:

Fit score calculation (example): - Account: Mid-market SaaS, $30M ARR, US-based, uses Salesforce + HubSpot, 25% YoY growth, Series A, hiring marketing - Industry (software): 100 * 0.20 = 20 - Company Size ($30M): 100 * 0.15 = 15 - Location (US): 100 * 0.10 = 10 - Tech Stack (both): 100 * 0.15 = 15 - Growth Rate (25%): 50 * 0.10 = 5 - Funding (Series A): 50 * 0.10 = 5 - Job Openings (marketing): 100 * 0.10 = 10 - Total Fit Score: 80/100

Accounts scoring 70+ on fit get on your target list. Accounts 60-70 are exploratory. Below 60, pass.

Engagement Score (Behavioral - Dynamic)

How actively is the account engaging with your company right now?

Engagement activities and weights:

Engagement score decay: - Points decay by 10% per week of inactivity - After 60 days of no activity, score resets to 0 - This encourages focus on active, engaged accounts

Example account engagement trajectory:

---

Fit + Engagement Combined Score

Overall Account Score = (Fit Score * 0.4) + (Engagement Score * 0.6)

This weights recent behavior (60%) more heavily than firmographic fit (40%), reflecting that active accounts are better prospects than perfect-fit but dormant accounts.

Score interpretation:

---

Implementation: From Theory to Practice

Step 1: Set Up in Your CRM

HubSpot implementation:

1. Create three custom properties: - Fit Score (number, 0-100) - Engagement Score (number, 0-100) - Overall Account Score (formula field)

2. Set up formula for Overall Score: (Fit_Score * 0.4) + (Engagement_Score * 0.6)

3. Create properties for each fit attribute: - Industry (dropdown) - Company Size (dropdown) - Location (text) - Tech Stack (multi-select) - Growth Rate (percent) - Funding Status (dropdown) - Hiring activity (checkbox or multi-select)

Step 2: Automate Engagement Score Updates

Trigger events to update Engagement Score:

1. Email opens: Workflow adds 1 point
2. Email clicks: Workflow adds 3 points
3. Website page visits (tracked): Add 2-8 points based on page type
4. Form submissions: Add 15 points, alert sales
5. Document downloads: Add 5 points

Decay automation: - Workflow runs weekly - Reduces Engagement Score by 10% for accounts with no recent activity - Resets to 0 if no activity for 60 days

Step 3: Import Fit Score Data

For new accounts or bulk import:

1. Pull Fit attribute data from ZoomInfo, Apollo, or Hunter
2. Map to your dropdown/select fields
3. Bulk-update Fit Scores (formula auto-calculates)
4. Use same tool quarterly to refresh company metadata

---

Testing Your Model: Validation Approach

Step 1: Validate Against Historical Data

Take 20 accounts that became customers in the past year and score them retroactively:

1. Score them with your new model
2. Compare to accounts that were opportunities but didn't close
3. If winners consistently score 60+, model has predictive power

Example validation: - Closed customers: Average score 72 - Lost opportunities: Average score 43 - Gap of 30+ points = strong signal your model works

If no clear gap, adjust weights or add new attributes.

Step 2: Test with Pilot Segment

Launch with 50 accounts at 70+ score:

1. Run for 30 days
2. Track: How many convert to opportunities?
3. Track: Sales reps' feedback on lead quality
4. Calculate: Cost-per-opportunity from this segment

Success criteria: - Opportunity rate: 15%+ - Sales feedback: "These are qualified" - Cost-per-opp: Lower than other sources

Step 3: Refine Weights Monthly

Review after 30, 60, and 90 days:

Adjust Fit Score if: - Many accounts with high Industry score don't convert (reduce industry weight) - All wins are from accounts with specific attribute (increase that weight) - New attribute emerges as predictive (add it)

Adjust Engagement Score if: - Certain activities don't correlate with sales conversations (reduce points) - Some activities predict later purchases even if low recent engagement (increase persistence weight)

Example refinement: - Problem: "Webinar attendance" is assigned 10 points, but 80% of attendees don't advance. - Solution: Reduce to 3 points. Or require engagement + webinar attendance for credit.

---

Common Model Mistakes

1. Over-weighting single firmographic attributes Don't let company size alone drive 40% of fit score. Use multiple attributes.

2. Setting engagement point decay too high If you decay 30% per week, a great lead from month 2 is worthless by month 3. That's wrong. Use 10% weekly.

3. Treating all form submissions equally A "Try Free" form is weaker than a "Schedule a Demo" form. Weight differently.

4. Ignoring negative signals If an account is heavily engaging with your competitor, they're active but deprioritize. Add a "competitor mention" field to reduce fit score by 20.

5. Never reviewing the model Set a quarterly review. Every quarter, 10-20% of your assumptions probably shift. Update them.

---

Advanced Scoring: Adding Negative Signals

Create a "Disqualification Score" to remove false positives.

Negative signals: - Company is direct competitor (disqualify) - Company is non-profit or education (if B2B SaaS focus) - Company is in consolidation/layoff mode (flag but don't disqualify) - Location is embargoed region (disqualify)

Implementation: - If disqualification signal exists, set Overall Score to 0 - If warning signal exists, reduce Overall Score by 30 points

---

Maintaining Your Model Over Time

Monthly Review (30 min, marketing + sales)

1. Review 3-5 accounts scored 70+ that didn't convert. Why not?
2. Review 3-5 accounts scored <50 that surprisingly converted. What missed?
3. Adjust weights if pattern emerges

Quarterly Deep Dive (2 hours, extended team)

1. Compare model performance: accounts by score band + conversion rate
2. Validate assumptions: Is fit still 40% vs. 60% engagement?
3. Update firmographic data (company sizes change, funding status changes, tech stack evolves)
4. Test new attributes (if industry data improved, add it)

Annual Refresh (half day, off-site)

1. Review full year of data
2. Rebuild model from scratch using validation data
3. Make major weight changes
4. Present to leadership

---

Actionable Checklist

• [ ] Define your Fit Score attributes and weights
• [ ] Assign weights to Fit attributes based on historical conversions
• [ ] Create Engagement Score activity list with point values
• [ ] Set up three custom properties in CRM (Fit, Engagement, Overall)
• [ ] Create formula field for Overall Score
• [ ] Audit 20 historical customers; validate model shows higher scores
• [ ] Identify and fix low-confidence attributes (replace or reweight)
• [ ] Automate engagement score updates via CRM workflows
• [ ] Import fit score data from data provider (quarterly refresh)
• [ ] Run pilot with 50 accounts at 70+ score
• [ ] Track conversion rate and cost-per-opportunity from pilot
• [ ] Refine weights after 30, 60, 90 days of pilot data
• [ ] Schedule quarterly model review meetings
• [ ] Document all assumptions and weights in shared doc

---

Expert Tips

1. Start simple, get complex Build a 5-attribute fit model first. Add more attributes only if they improve predictive power.

2. Engagement decay matters more than point values Getting the decay curve right (weekly 10% vs. no decay) matters more than whether a form is worth 15 vs. 20 points.

3. Validate against real conversions Your model is only as good as its correlation with closed deals. If you haven't validated, it's just a hypothesis.

4. Use negative scores for disqualification Don't include disqualifiers as negative points in overall score. Use them as veto gates: if disqualified, score = 0.

5. Keep the model transparent Sales and marketing should understand why an account got its score. If they don't, they won't trust it.

---

Skip the 9-tool stack. Book a 30-min Abmatic AI demo ->