Lead scoring model

A lead scoring model is the set of rules that turns a lead’s attributes and behaviour into a single number. It does not decide the handoff. A marketing qualified lead is the threshold drawn on the score; nurturing is what happens below it.

The point of a model is comparability. Two leads from different campaigns, different months and different sources can be ranked against each other because the same rules were applied to both.

That ranking is what makes a lead qualification rate meaningful. Without a shared model, the rate just measures how optimistic whoever filled in the form field happened to be feeling.

Key takeaways

• Scoring assigns the number; the MQL definition draws the line — nurturing works everything underneath it.
• Score fit and behaviour separately, then combine them, so a poor fit cannot buy its way up on clicks.
• Decay matters as much as accrual — a score with no decay turns into a seniority list.
• A score is a diagnostic, not a target: do not report it as a performance measure.

How it works

Most models score two things separately and then combine them. Fit covers who the lead is, such as job title, industry and company size. Behaviour covers what the lead did, such as pages viewed, webinars attended and emails opened.

Oracle’s documentation for its marketing automation product sets out the pattern. Profile criteria are explicit data such as title, industry or company revenue. Engagement criteria are implicit data such as site visits.

In that product the two combine into a letter and a number, so A1 is the most qualified lead, D4 the least.

The point values are arguments — not facts. Set them in a room with sales, then test them against last year’s closed deals to see whether the model would have ranked the winners near the top.

Decay is the part most teams skip. Interest ages, so a lead that read three pages in March and nothing since should not outrank a lead that read two pages last week.

The score lands in a customer relationship management (CRM) record, where it becomes visible to the people who act on it. Keeping it in the marketing tool only is how models quietly stop being used.

Scores invite misuse. Google says plainly that its advertising Quality Score is a diagnostic tool, not a key performance indicator and should not be aggregated with the rest of your data. The same caution applies to a lead score.

A sales development representative should read a score as a work queue order, not as a promise. The number says who to call first, not who will buy.

Keep the model small. A model with forty signals cannot be explained to the sales floor, and a rule nobody can explain is a rule nobody trusts the first time the ranking looks wrong.

The model stops at the number. Where the line goes, who receives the lead and how fast they respond all belong to the marketing qualified lead definition, which is a separate agreement between two teams.

Review the model quarterly against closed-won data. If high scorers close at the same rate as low scorers, the weights are wrong and no amount of extra signals will fix them.

Examples

Scoring models differ mainly by how much behaviour they can see. A high volume business with heavy web traffic can afford a behavioural model; a business with twenty target accounts cannot, and should score fit almost exclusively.

A business software vendor scores fit from the CRM system and behaviour from the website, then routes anything above 75 points straight to a sales queue. Everything else returns to a nurture track with the score visible.

An outsourced sales team scores differently again. When the calling desk sits offshore, the model has to encode local knowledge the team does not have, so industry, company size and region carry more weight than interest signals.

A subscription business ties scoring to cost per lead by source. Channels producing high scoring leads at a low cost get more budget the following quarter, which is the only reliable use of the score outside the sales queue.

Fit and behaviour answer different questions — who this is, and how interested they are. A model that merges them into one number too early loses the distinction that makes routing possible.

Related terms

• Lead qualification rate: the share of leads that clear the bar; the model is what sets the bar consistently.
• Customer relationship management (CRM): the record the score is written to and acted on from.
• Sales pipeline: the stages a scored lead moves through once sales accepts it.
• Pipeline velocity: how fast value moves through those stages; better scoring usually shows up here first.
• Cost per lead: what a lead cost to acquire, which only means something once scoring tells you its quality.

FAQ

What is the difference between a lead scoring model and an MQL?

The model produces a number for every lead. The marketing qualified lead definition is the single threshold drawn on that number, plus the agreement about what sales does when it is crossed.

How many points should each signal be worth?

There is no standard set. Start from the behaviour of deals you actually closed last year, assign weights that would have ranked those deals highly, then adjust quarterly.

Should the model include negative points?

Yes. Free email domains, competitor traffic, student job titles and long gaps in activity all need to pull a score down, or the number only ever rises.

Can one model serve the whole business?

Rarely. Separate product lines, regions or deal sizes usually need separate models, because a signal that predicts a small purchase often says nothing about a large one.

Who owns the model, marketing or sales?

Both, which is why it needs a scheduled review rather than an owner. Marketing maintains the mechanics and sales confirms the weights still match what closes.

Compare vetted demand generation partners across the Outsource Accelerator network before you build the model with an agency.