Lead Scoring

Lead scoring is the method of assigning a numerical score to each lead based on how well they fit your ideal customer and how interested they seem. The idea is simple: instead of treating every contact the same, the sales team prioritizes the ones who accumulate the most points, because statistically they are the ones closest to buying.

The score is built by adding (and sometimes subtracting) points for different signals. An operations manager at a large company who downloads a case study scores more than a student who opens an email once. That way, the team invests its time where there is the highest chance of closing, instead of chasing cold contacts.

In an operation running on Salesforce, lead scoring lives inside the CRM and feeds the prioritization in Sellium, where each lead arrives sorted by its score before a sales rep works it.

The two dimensions of the score: attributes vs behavior

A good lead scoring model combines two types of signals that should not be confused. On one side are attributes (demographic or firmographic scoring): who the contact is and which company they belong to. Role, industry, company size, country, revenue. These signals measure fit (how closely they resemble your ideal customer profile). On the other side is behavior: what the contact does. Website visits, downloads, email opens, clicks, demo requests, attending a webinar. These signals measure interest and buying intent over time.

The difference is key: a contact can have perfect fit but zero interest (the ideal role who never opened an email) or a lot of interest but little fit (a student who devours all your content but will never buy). The most useful models separate the two axes and only flag as a priority the lead that scores high on both. Those with fit but little interest are sent to lead nurturing; those with a lot of interest and little fit are left to mature or discarded.

Attributes vs behavior, side by side

Example of a simple points model

Imagine a B2B company in Argentina that sells business management software. Its model could look like this, with a 50-point threshold for passing a lead to sales:

• Management or executive role: +20
• Company with more than 50 employees: +15
• Target industry (for example, Consumer Goods): +10
• Corporate email (not Gmail or Hotmail): +5
• Requested a demo: +25
• Visited the pricing page: +15
• Downloaded an ebook: +10
• Opened three or more emails in the week: +5
• Free personal email: -10
• No activity in 30 days: -15

With that table, a manager at a small consumer goods company who requests a demo and looks at pricing quickly reaches the threshold and shows up at the top of the rep's queue. A contact with a Gmail address who only opened one email stays far from the cutoff and remains in an automated nurturing sequence.

Why it matters in Latin America

In small sales teams, which are common at many companies in the region, the rep's time is the scarcest resource. Without scoring, the salesperson works in order of arrival or by intuition, and ends up spending hours on contacts who were never going to buy while a hot lead goes cold waiting. A points model, even a basic one, turns a messy list into a prioritized queue and makes follow-up consistent, not dependent on the mood of the day.

Common mistakes

The first mistake is overcomplicating the model from the start: launching with 40 rules and finely tuned weights without historical data. It is better to start simple, with five or six clear signals, and adjust later. The second is never subtracting points: if a lead shows no activity for weeks or gives a throwaway email, the score should go down (negative scoring or decay), because interest expires. The third is not closing the loop with sales: if the sales team says those 80-point leads were no good, you have to review the model, not ignore the data. Scoring is a hypothesis that gets calibrated against real closing results.

Many CRMs let you go beyond the manual rules model and use predictive scoring, where a machine learning model analyzes your past won and lost opportunities to discover on its own which signals predict a close, and assigns the score without you defining each weight by hand.