Lead Scoring

Without lead scoring, your sales team works leads in the order they came in. First in, first called. That's like a hospital treating patients in the order they arrived instead of by severity. The person with a paper cut gets seen before the person having a heart attack.

Lead scoring is triage for your pipeline. It tells reps: "This lead visited pricing three times, works at a 500-person company in your target vertical, and just started a chat — call them now." Meanwhile, the free email address who downloaded one ebook three weeks ago can wait.

The numbers back this up. Organizations with lead scoring see 77% more lead generation ROI than those without, according to Lenskold Group research. Not because they generate more leads, but because they stop wasting sales time on the wrong ones. Salespeak.ai's AI qualification layer adds real-time conversational scoring on top of traditional behavioral signals — it doesn't just track what people do, it talks to them and scores based on what they say.

Lead scoring models typically combine two types of signals:

1. Demographic/firmographic scoring (fit): Does this lead match your ICP? Company size, industry, job title, location, tech stack — each attribute gets a weight. VP at a 200-person SaaS company? +25. Student email? -15. This tells you who the lead is.
2. Behavioral scoring (interest): What has the lead done? Visited the pricing page (+20), opened 5 emails in a week (+15), attended a webinar (+10), unsubscribed (-30). This tells you how engaged they are.
3. Intent scoring (timing): Third-party intent signals — are they researching your category on G2, Gartner, or other review sites? Are they hiring for roles that suggest they'll need your product? This tells you when they might buy.
4. Score threshold and routing: When a lead crosses a predefined threshold (say, 80/100), it's automatically routed to sales as an SQL. Below that threshold, it stays in marketing nurture. The threshold should be calibrated to your actual conversion data.
5. AI enhancement: Modern AI scoring analyzes all three signal types simultaneously, plus conversational data from chat interactions. It finds patterns humans miss and updates dynamically. A lead who just told the AI "We need this live by Q2" gets a massive score boost in real time.

An HR tech company had 1,200 leads per month flowing into their CRM. Eight sales reps worked them in order of recency. The result was chaos — reps spent 20 minutes on calls with 5-person startups using free Gmail accounts, while enterprise prospects from target accounts sat unworked for 48 hours.

They built a scoring model. Simple rules to start: +20 for company size over 100, +15 for HR director or VP title, +25 for pricing page visit, +30 for demo request, -20 for student email or non-target industry. Leads scoring above 70 got routed to a senior AE. Leads 40-70 went into nurture. Under 40 got a self-serve path.

Three months later, they layered in AI scoring that analyzed which combinations of behaviors predicted closed-won deals. The AI found that leads who visited the integrations page AND had a company page on Glassdoor with open HR roles closed at 5x the rate. Nobody would've guessed that pattern manually.

Pipeline velocity increased 34%. Average deal size went up 18% because reps were focusing on bigger, better-fit accounts. One rep said: "I used to dread my call list. Now I'm actually excited to make calls because I know these people want to talk to me."

• Scoring based on gut feel instead of data. "I think whitepaper downloads should be worth 20 points" is a guess. Look at your last 200 closed-won deals. Which behaviors did they share? Score based on what actually predicts conversion, not what feels important.
• Never decaying scores. A lead who visited your site 8 months ago isn't the same as one who visited yesterday. Scores should decay over time. If someone hasn't engaged in 30 days, their behavioral score should drop. Recency matters more than total activity.
• Marketing and sales disagreeing on what a "good" score means. If marketing passes leads at 50 and sales wants them at 80, you've got an alignment problem. Get both teams in a room, agree on the threshold, and revisit it quarterly.
• Too many scoring rules. Start with 5-10 rules. A model with 47 scoring criteria is impossible to debug when it stops working. Simple models you understand beat complex models nobody maintains.
• Ignoring negative scoring. Don't just add points — subtract them. Competitor employee? -50. Unsubscribed from emails? -20. Job title is "student"? -30. Negative scoring prevents bad leads from accidentally reaching the threshold.