Glossary
What is lead scoring?
Lead scoring is a way of ranking leads by how likely they are to buy, by scoring what you know about the company (fit) and what the company has recently done (intent), then working the list from the top.
Fit and intent measure different things
Fit is about the account: industry, size, location, technology, whether they look like the customers you already keep. It barely changes month to month, and it is knowable before anyone talks to you.
Intent is about behaviour: pages read, emails opened, a demo booked, a reply that took two minutes. It changes constantly and it decays fast. A visit from last April tells you nothing about this week.
Scoring them together in one number is where most models go wrong. A perfect-fit company doing nothing and a poor-fit company doing everything can land on the same total, and they need completely different treatment. Keep the two scores visible separately, even if you rank on the sum.
Points models and predictive models
A points model is a set of rules a person wrote: plus fifteen for the right industry, plus ten for a pricing page visit, minus twenty for a free email domain. It is easy to explain, easy to argue with, and easy to fix. That is its main advantage, and it is a bigger advantage than it sounds.
A predictive model learns weights from your own closed-won and closed-lost history. It is better when there is enough history to learn from, which in practice means hundreds of closed deals, not dozens. Below that it is fitting noise, and it will tell you so with great confidence.
Either way, the model needs a review schedule. A scoring model that has not been checked against outcomes in a year is a set of assumptions from a year ago.
The negative signals matter more than people expect
Most models only add points. Subtracting is what keeps the top of the list clean: a competitor domain, a student email address, a job applicant, a country you do not serve, a role that never buys.
Without negative scoring the top of the list slowly fills with things that score well and cannot buy, and the team stops trusting the ranking, which is the end of the model whether anyone turns it off or not.
How lead scoring works in a CRM
- A visible score, not a hidden one. If a rep cannot see why a lead scores 82, the score is a number they are asked to obey. Show the components.
- Decay on intent. Behaviour scores should fall over weeks. Without decay, a lead that was hot in March stays at the top forever.
- A threshold that triggers something. A score with no action attached is decoration. Crossing the line should route the lead, notify the owner, or start a sequence.
- Outcome tracking. Store the score at the time of the win or the loss. That is the only way to find out later whether the model was any good.
How we do it in Lumenbase
LumenScore is our version of this. It combines fit signals from the company record with engagement the system captured on its own, from email, meetings, and LinkedIn, so the intent half is not waiting on anyone to log an activity. The components are shown on the record rather than rolled into one opaque figure.