B2B lead scoring: how to build a truly useful score?

What is lead scoring?

Lead scoring assigns a value to a prospect based on the information available in your CRM: company profile, the contact's role, the expressed need, the project's timeline, or even certain behavioral signals.

However, it does not replace sales qualification.

Qualification seeks to determine if a prospect actually deserves to move forward in the sales process. Scoring, on the other hand, allows certain criteria to be made measurable and reproducible in order to rank or prioritize leads.

If this first step is not yet structured, it is generally better to start by defining how to properly qualify B2B leads before building a scoring system.

Scoring primarily gains value when it is part of a larger Revenue process: acquisition, qualification, opportunity, conversion, and then expansion. This is one of the topics we detail in our article dedicated to RevOps.

What is lead scoring actually used for?

Lead scoring becomes interesting when it allows for a concrete decision to be made.

In particular, it can be used to:

1. prioritize leads when not all of them can be handled immediately;

2. automatically route a prospect to the right salesperson or team;

3. trigger an action: task, notification, nurturing sequence, or follow-up;

4. avoid engaging Sales too early on prospects who are not yet mature enough.

The main rule to keep in mind is simple:

a score only has value if it changes a decision or an action.

Displaying "Lead score: 76" in a CRM without knowing what should happen at 76 rather than 52 ultimately brings little value.

Do you really need lead scoring?

Not every B2B company needs a scoring system.

An agency that receives fifteen inbound requests per month and can analyze each of them individually is probably better off structuring a few qualification fields than maintaining a model made of twenty criteria.

For example:

• need;

• company type;

• order of magnitude of the project;

• deadline;

• next action.

Lead scoring becomes more relevant as complexity increases:

• multiple channels generate leads;

• not all prospects can be handled immediately;

• several salespeople must share the opportunities;

• certain actions can be automated;

• enough data is available to actually differentiate prospects.

However, there is no universal threshold above which a company should implement lead scoring.

The right question is rather:

Do we really need a mechanism to automatically decide which prospects deserve our attention?

If the answer is no, a simple and properly used qualification process will often be preferable.

Fit and intent: the two dimensions to distinguish

One of the limitations of a single score is that it easily mixes information that does not answer the same question.

It is often more useful to distinguish fit and intent.

Fit: is this a prospect we want?

Fit measures how well the prospect matches the type of customer the company actually wants to support.

Depending on your business, it can take into account:

• how well the need matches your offer;

• the type of company;

• its size;

• its sector, when this is highly discriminating;

• its location;

• the role of the contact.

Not all of these criteria are relevant everywhere.

A company specializing in a specific sector might give a lot of weight to the prospect's industry. Another might serve an agency, a consulting firm, or a startup in the same way: the sector should then barely influence its score.

Intent: does the prospect seem ready to move forward?

Intent seeks to determine if the prospect is currently showing purchase intent or sufficient maturity.

The most interesting signals are often the most direct ones:

• request for a meeting or demonstration;

• clearly expressed problem;

• already identified project;

• relatively close deadline;

• interaction with a commercial proposal;

• usage signals in a SaaS context.

Marketing signals must be interpreted with more caution.

Opening a newsletter, downloading a resource, or viewing multiple articles shows engagement. It does not automatically mean the person wants to buy.

A prospect who specifies they want to replace their CRM within the next three months generally sends a much stronger signal than someone who has opened five emails.

The Fit × Intent Matrix

Rather than immediately converting all this information into a score out of 100, a simple matrix already helps in making useful decisions:

Now let's imagine two leads:

Lead A

• Fit: 9/10

• Intent: 5/10

Lead B

• Fit: 4/10

• Intent: 10/10

Both get 14 points if you simply add up their scores.

Yet they tell a different story.

The first matches your target very well but may not be ready to buy yet. The second seems to want to move forward quickly, but does not necessarily match the type of customer you want to sign.

The associated actions must therefore be different.

How to build your lead scoring model?

A good model does not start with a point grid found on the Internet.

It starts with the decision you are trying to improve.

1. Start with the decision to be made

First, ask the question:

What do we want to do differently because of the score?

For example:

• call certain leads with priority;

• automatically assign requests;

• send certain prospects into a nurturing sequence;

• trigger a human qualification.

If no action is associated with the result, scoring risks simply becoming just another metric in the CRM.

2. Identify a few highly discriminating criteria

Look at your own data and your teams' experience:

• which profiles actually become customers?

• which opportunities are regularly lost?

• which criteria do salespeople already use to prioritize?

• which signals actually precede a conversion?

With little history, five relevant criteria will often be more useful than a model with thirty variables that gives an illusion of precision.

3. Separate fit and intent

A stable characteristic of the company does not necessarily need to be mixed with a one-off behavior.

A company's size describes its fit.

A demonstration request describes its intent instead.

Keeping this distinction makes it easier to both understand and evolve the system.

4. Weight without seeking false precision

Not all criteria are equally important.

A need that matches your offering perfectly can count more than a secondary characteristic.

But assigning +4 rather than +3 is not a scientific truth. Weightings primarily represent a hierarchy, which must then be tested against real results.

5. Disqualify rather than subtract dozens of points

Not everything should be turned into a score.

If a characteristic actually makes a prospect incompatible with your offering, an explicit rule is sometimes better:

Company located in an unserved area → disqualified

rather than:

Unserved area → -50 points

The score can then be used to differentiate only the prospects who actually have a chance of becoming customers.

6. Associate each level with an action

For example:

• High Fit + High Intent → priority sales handling;

• High Fit + Low Intent → follow-up or nurturing;

• Low Fit + High Intent → quick qualification;

• Low Fit + Low Intent → no immediate sales action.

The scoring then becomes a genuine operational mechanism.

7. Test then adjust

Test your model on past leads.

Would it have surfaced the opportunities that were ultimately won? Do certain criteria carry too much weight? Do the sales representatives find the results consistent?

Scoring must evolve along with the ICP, offers, and acquisition channels.

Simple B2B Lead Scoring Example

Let's take a B2B company wishing to prioritize its inbound requests.

It could start with two scores out of 10.

Fit Score

Intent Score

This grid is deliberately not a template to be copied as is.

For one company, size will be decisive. For another, almost irrelevant. You should only keep the criteria that actually change the likelihood of a prospect being relevant or ready to move forward.

Mistakes that make lead scoring useless

Adding too many criteria

Each new criterion increases the required data and model maintenance. A score based on twenty rarely populated pieces of information will be less reliable than a simple model fed properly.

Giving points to every interaction

Not all interactions are buying signals. An email open or an article visit should not automatically turn a prospect into a sales priority.

Confusing fit and engagement

An out-of-target prospect can consume a lot of content. Conversely, a company perfectly suited to your offer may still have interacted very little with you.

Using unreliable data

Scoring automates a decision based on your data. If your data is incomplete or poorly structured, the score will mostly automate its errors.

Never challenging the model

Your ICP, offers, and channels evolve. Scoring must be regularly tested against the opportunities actually created and won.

How to integrate lead scoring into your CRM?

In many cases, no complex architecture is necessary.

A simple model can work like this:

CRM data → formulas → scores → workflow → sales action

If your CRM can already store criteria, calculate the result, and trigger the necessary actions, adding Make, n8n, or a backend solely to calculate a few points rarely brings value.

The main challenge remains the structuring of the CRM: reliable data, an understandable model, and actions actually adopted by the teams. This is the logic that should guide a CRM implementation or redesign.

Example with Attio

Attio now allows setting up this type of logic directly within the CRM.

Its Formula attributes allow calculating values from existing data. The official documentation features an example of lead scoring combining fit and intent, while workflows then allow using this information to trigger appropriate actions.

For a relatively simple scoring setup, staying within Attio therefore avoids automatically adding an external automation layer.

A complementary architecture primarily becomes relevant when the score depends on data coming from a SaaS product, a data warehouse, or multiple external systems.

What role for AI in lead scoring?

AI can improve certain steps of scoring, but it should not unnecessarily replace simple rules.

For deterministic information—company size, country, account type—a formula generally remains sufficient.

AI becomes more interesting when the data is unstructured. For example, it can analyze a free-form response to a form to identify the need, timeline, or certain context elements, and then convert this information into actionable attributes for the CRM.

This is also a logic found with Attio's AI and MCP features, which notably allow structuring or better leveraging the information present in the CRM.

Finally, when a company has a lot of history and volume, predictive scoring can look for correlations that are difficult to formalize manually.

But if five understandable rules already allow making the decision properly, adding AI does not automatically make the system better.

Lead scoring, MQL and SQL: what is the link?

Lead scoring can help move a prospect along the marketing and sales cycle.

A sufficient level of fit and intent can, for example, trigger sales reps to take over.

But an MQL or an SQL should not be defined solely by an arbitrary threshold like "score higher than 70". The score remains one signal among the rules used to determine the prospect's actual maturity and qualification.

Frequently asked questions about lead scoring