A different problem from pipeline scoring

Two things get called scoring and they solve different problems.

Pipeline scoring ranks deals already being worked — opportunities with a value, a contact and some history.

Lead scoring happens earlier. An enquiry has just arrived. Nobody has spoken to them. The question is: is this worth a call, and how soon?

This article is about the second one. If your problem is prioritising open deals, that is a different exercise.

The problem it solves

Most businesses that market at all get more enquiries than they can properly follow up.

Some are serious buyers. Some are students doing research. Some are competitors. Some are people who downloaded a guide and have no intention of buying anything.

Without a filter, one of two things happens. Either everything gets called, which wastes your sales team’s time. Or the newest ones get called and older enquiries quietly rot, including good ones.

Lead scoring is a triage tool. It decides what gets attention first.

What it looks at

Two categories, and both matter.

Who they are

• Company size and industry
• Job title of the contact
• Country or region
• Whether they are an existing customer
• Whether their company matches your typical buyer

What they did

• Which pages they looked at
• Whether they returned more than once
• What they downloaded
• Whether they opened your emails
• How they found you — search, referral, campaign, event

The second category is usually more predictive than the first. A junior person from a small company who has read your pricing page three times is often a better lead than a director who filled in a form once and never came back.

Behaviour is intent. Profile is only eligibility.

The requirement, again

You need enough converted leads to learn from.

A model trained on a few dozen conversions is not learning your business. It is finding coincidences in a small sample and presenting them with confidence.

Two other things matter as much:

You must record what happened to leads that did not convert. If unqualified leads are deleted or left sitting forever, half the training data is missing. The model needs to see failures.

Your lead source data must be accurate. If half your leads are tagged “website” because nobody set up proper source tracking, the strongest available signal is unusable.

Where it helps

Ordering the follow-up queue. The main use. Which five enquiries deserve a call this morning.

Speed on the good ones. Response time matters enormously in most markets. A leadscored highly and called within the hour converts far better than the same lead called in three days.

Routing. Sending enterprise-shaped enquiries to a senior salesperson and smaller ones to a lighter-touch process.

Deciding what to nurture rather than call. A lead that is not ready is not a bad lead. Scoring helps separate “not now” from “not ever”.

Where it does not

Deciding who to ignore entirely. A low score means “less like your past conversions”. Some of your best customers will not look like your past customers.

Judging fit. A score cannot tell you whether you can actually serve this customer well. That is a human question and sometimes the more important one.

Working without volume. If you get thirty leads a month, sort them yourself. The overhead of a scoring model exceeds the benefit.

The trap, once more

The same self-fulfilling problem appears here as in pipeline scoring, and it is worth repeating because it is easy to miss.

Low-scored leads get no attention. They fail to convert — because nobody called them. The model observes this and becomes more confident that leads like that do not convert.

The score makes itself right.

Two safeguards:

Work a deliberate sample of low-scored leads. Not all of them. Enough to keep the data honest and occasionally find something.

Check calibration quarterly. Of the leads scored highly, what proportion converted? Of the low-scored ones you did work, what proportion converted? If those numbers do not match what the score implied, it is not learning your business properly.

What actually moves the needle

An uncomfortable point, and an important one.

For most businesses, response speed matters more than scoring accuracy.

A mediocre score with a one-hour response will beat an excellent score with a three-day response, in almost every market. Leads go cold fast, and the first credible response often wins.

If you are choosing where to put effort, fix response time first. Scoring helps you decide the order within an hour. It does not help if the hour is actually a week.

Setting it up

1. Fix source tracking. Know where leads actually come from. This is the strongest single signal and it is frequently broken.

2. Record outcomes properly. Converted, or not converted with a reason. Both are training data.

3. Measure your baseline. Conversion rate by source, and average response time.

4. Run it silently for a quarter. Score leads without showing anyone. Then check whether high-scored leads actually converted more.

5. Present it as an order, not a number. A percentage next to a company name invites false confidence and arguments. “Suggested priority” is honest about what it is.

6. Keep working some low scores. Permanently.

Measuring it

Conversion rate. Are more of the leads you work converting?

Response time on high-value leads. Getting to good leads faster is the main mechanism.

Wasted calls. Time spent on enquiries that were never going to buy.

Leads left untouched. Watch this. If it grows, you are not triaging — you are abandoning.

The short version

Lead scoring is triage. Used to order a follow-up queue, it saves real time and gets your team to good enquiries faster.

It needs volume, recorded outcomes, and accurate source data. Without those it is a confident guess.

And before investing in it: check your response time. For most businesses that is the bigger lever, and no amount of scoring compensates for calling three days late.