What it means
Lead scoring is the practice of attaching a number to every contact in your CRM that summarises how likely they are to buy. The score is a function of two things: fit (does this person match your ideal customer profile?) and intent (are they actively showing buying behaviour?). Most teams blend the two into a single number from 0-100.
The point is not the number itself; the point is to rank a list of 5,000 leads so that your three SDRs work the right 200 first tomorrow morning, instead of taking them top to bottom by timestamp.
Two flavours exist:
- Rules-based scoring assigns explicit point values to specific signals: "+30 for filling a form, +20 for visiting the pricing page, -15 for using a personal email address". The model is transparent and easy to explain to a skeptical sales team.
- Predictive scoring uses a model trained on your historical closed deals, typically logistic regression or a gradient boost, to predict the conversion probability of each new lead. More accurate at scale, but a black box if you do not show feature importance to the team.
Most successful programmes start with rules-based and graduate to predictive once they have at least 1,000 closed deals and a quarter of stable rules.
Demographic vs behavioral scoring
The cleanest mental model is to split the score into two components and combine them at the end:
- Demographic / firmographic score (fit). Job title, seniority, company size, industry, geography, tech stack. This score answers: "If they bought, would they be a good customer?"
- Behavioral score (intent). Pages visited, forms filled, demo requested, content downloaded, DM replies, webinar attendance. This score answers: "Are they showing up?"
Many teams plot the two on a 2x2:
- High fit + high intent → call within an hour. These are your SQLs.
- High fit + low intent → automate nurture. Drip campaigns, retargeting ads.
- Low fit + high intent → polite redirect. Self-serve plan, help docs, community.
- Low fit + low intent → suppress. Do not waste rep time. Let decay rules drop them off the list.
Sample weighting (and where it comes from)
Below is the kind of weighting a SaaS or D2C team running on CRM Solid would start with. Adjust to your funnel; if you do not have a starting point, this is a fine one.
Positive signals (behavioral and demographic):
- Form fill on contact page: +30
- Pricing page visit (>15s dwell): +20
- Demo request: +50
- Free-trial signup: +60
- Webinar attended: +25
- Title contains "founder", "CEO", "head of": +25
- Company size 20-500 employees: +20
- In target geography: +15
Negative signals:
- Personal-email domain (gmail, yahoo, outlook): -15
- Title contains "student", "intern": -20
- Unsubscribed from email: -30
- No activity in 30 days: -40 (decay)
Threshold the result so that any contact with a score above 60 becomes an MQL, and any contact above 90 is an SQL routed directly to a rep. Reset every 90 days and recalculate.
Decay rules
Without decay, your lead-scoring model is a permanent inflation machine. Every action adds points, nothing subtracts them, and after a year your top-scored leads are the ones who clicked an email in 2024, not the ones who clicked yesterday.
A standard decay schedule:
- 7 days idle: -10
- 14 days idle: -25
- 30 days idle: -40
- 90 days idle: route to long-term nurture, freeze score
"Idle" should include all channels you measure, not just email. A contact who replied on Telegram yesterday is not idle.
Why it matters
Three reasons. Rep efficiency: a focused list is worth 3-5x a chronological list. Conversion rate: calling a high-scored lead within an hour is worth a measurable multiple in close rate; the Harvard Business Review's often-cited "lead response time" study put it at 7x for the first hour. Marketing accountability: a transparent scoring model is the cleanest way to debate "are these leads actually qualified?" with marketing; you can argue about weights, not about gut feel.
Real-world examples
- A B2B SaaS team scores every signup on fit (company size, role) and intent (pricing-page visits, demo requests). Scores above 80 are auto-routed to AEs; 50-80 go to SDR call queues; under 50 get a 6-week nurture drip.
- An agency scores every contact-form submission by industry match, project budget (from a form question) and page-visit history. The owner only sees leads above the SQL threshold; everything below routes to a junior account exec for triage.
- A real-estate agent uses scoring to rank Instagram-DM leads: budget mentioned in DM (+25), property type specified (+15), city match (+20), already pre-approved for a loan (+50). Hot leads get a phone call within 30 minutes; cold leads get a Telegram nurture sequence.
- A SaaS startup running PLG scores in-product behavior more heavily than marketing-site visits. "Invited a teammate" is +40. "Connected a real data source" is +50. "Started a workflow" is +30. The model deprioritizes contacts who only ever visit pricing pages without using the trial.
- A creator selling on DMs scores Telegram followers by message volume, response time, and whether they asked a price question. A 90+ score triggers a hand-off from the AI agent to a human DM.
Common mistakes
- No negative weights. Without negatives, every new signal pushes the score up. Negative weights for personal emails, low-fit titles, and inactivity are what make the model actually rank, not just accumulate.
- No decay. A lead's hotness has a half-life. Inactivity has to subtract points or your top leads will all be a year old.
- Only scoring marketing-site behavior. The most valuable signals usually come from in-product actions, DM replies, sales calls, or live chat, not blog reads.
- Ignoring channel asymmetry. A DM reply is a higher-intent signal than an email open. Weight accordingly.
- Building the model alone in marketing. If sales does not believe the score, sales will not act on it. Build the model with at least one AE in the room.
Related concepts
- Sales pipeline: where scored leads land after handoff.
- Conversion funnel: the upstream funnel that feeds the score model.
- Drip campaign: what you do with leads who score too low for a call.
- Lead magnet: most of your "+30 form fill" signals start here.
- Omnichannel CRM: a prerequisite for scoring signals across every channel consistently.
- AI agent: can act on high-scored leads autonomously while a human handles the rest.
How CRM Solid handles it
CRM Solid ships with built-in lead scoring on the contact record: both demographic and behavioral signals, with weights you set in the UI. Decay rules run nightly. AI-agent and DM-reply signals feed the same model as email and form fills, which is the part most CRMs get wrong. Scoring data is available via the public API and the Contacts CRM page, so external tools (ad platforms, data warehouses) can use the same number.