AI Lead Scoring

Why lead scoring matters

At 100 inbound leads/month, a good salesperson can reach out to everyone. At 500 leads/month, you have to prioritize. At 5,000 leads/month, you need automation or you miss opportunities.

Lead scoring assigns each lead a number (typically 0-100) based on signals that correlate with closing. Your sales team works the top 20% first, and recovers the middle 60% via automated nurture.

What signals Monfri uses

We score on three dimensions:

1. Demographic fit (who they are)

• Company size (employees, revenue — via enrichment)
• Industry / vertical
• Role / seniority of the contact
• Geographic region
• Company funding stage (for B2B SaaS)

2. Behavioral engagement (what they do)

• Pages visited (pricing page = strong signal)
• Email engagement (opens, clicks on last 30 days)
• Demo scheduled, trial started, feature usage depth
• Recency — activity in last 7 days is weighted 3× vs 30 days ago

3. Firmographic + intent (context)

• Existing deal stage (if already in pipeline)
• Referral source (demo requests convert 3× inbound form fills)
• Multiple contacts from same company (buying committee forming)

How the model learns

We start with a rule-based baseline (reasonable weights you can adjust) and progressively tune based on your closed/lost deals.

Cold start (first 30 days)

You set weights manually: "pricing page visit = +15 points," "demo request = +30," "< 10 employees = -10." This is essentially manual scoring, which most teams already do in a spreadsheet.

Warm model (30-90 days)

After ~50 closed-won and 200 closed-lost deals, our model regresses on the factors that correlated with wins. You see weight recommendations: "Your actual win rate on pricing page visitors is 12% vs 3% baseline — suggest raising this signal's weight."

Mature model (90+ days)

At ~500 closed deals, the model handles weighting automatically. You inspect and override as needed. This is when "AI" is actually doing work vs theater.

Honest accuracy expectations

Lead scoring isn't magic. Realistic outcomes:

• Top-quartile leads (score 75+): 30-50% close within 60 days. Prioritize personally.
• Middle-half leads (score 25-75): 5-15% close. Automated nurture + light touch.
• Bottom-quartile leads (score <25): <3% close. Newsletter-only.

If a vendor promises "90% accurate lead scoring," ask what metric they're measuring. Real-world B2B SaaS lead scoring with good data gets you 40-60% improvement in conversion per sales rep hour, not order-of-magnitude.

Setting it up in Monfri

1. Turn on Lead Scoring: Settings → CRM → Lead Scoring
2. Review default weights (we ship industry benchmarks)
3. Add custom signals: any custom field or event you track can be a signal
4. Set thresholds: define "hot" (>70), "warm" (30-70), "cold" (<30)
5. Wire automation: "when lead reaches 70+, assign to Tier 1 salesperson" or "when lead drops below 20, enroll in re-engagement sequence"

Limitations worth knowing

• Data quality matters more than algorithm. Garbage in = garbage out. Invest in clean CRM data before sophisticated scoring.
• Humans overrule the model. Your salesperson knows things the model doesn't — let them override scores manually.
• Scoring isn't a replacement for judgment. Use it to prioritize, not to eliminate leads from consideration.
• Recalibrate quarterly. Buying patterns shift. Re-review model weights every 3 months.