In short
A lead is a person or business that has shown interest, for example by filling in a form, calling or downloading a guide. Lead scoring adds points for signs of a good fit, such as being in your service area or needing a service you offer, and for signs of intent, such as asking for a quote or visiting the pricing page. Points may be taken off for poor fit. The total decides who gets a call first and who goes into a slower follow-up sequence.
Fit and intent
| Fit | Does the lead match your ideal customer? Location, type and size of business, budget, the service they need. |
|---|---|
| Intent | How ready are they to buy? Requested a quote, booked a call, replied to an email, returned to the site. |
| Negative signals | Outside your area, a job type you do not take on, a student or job seeker, an unsubscribe. |
| Decay | Points that fade over time, so an enquiry from six months ago no longer outranks one from today. |
Rule-based versus AI scoring
Rule-based scoring uses points you choose. It is easy to explain and change, and it works well when you have a clear idea of what a good lead looks like. Its weakness is that the rules reflect your assumptions, which may be wrong.
AI or predictive scoring learns from your history: which past leads became customers and what they had in common. It can spot patterns people miss, but it needs a reasonable volume of past leads with recorded outcomes, and clean data. With only a few dozen enquiries a month, simple rules are usually the better place to start.
Language models add another option: reading the text of an enquiry or a call transcript and pulling out the job type, urgency and budget, which then feed the score. This turns free-text messages into fields you can sort by.
How to set it up
Look at your last 50 to 100 won and lost enquiries and note what the wins had in common. Turn the strongest three to five signals into rules, and set a threshold above which a person calls the lead the same day. Everyone below the threshold gets a useful automated follow-up instead of nothing.
Agree the score with the people who make the calls. If the team does not trust it, they will ignore it. Review the results after a month: are high-scoring leads converting better than low-scoring ones? Adjust the rules if not.
Make sure the data the score depends on is captured consistently. If postcode, job type or budget are only sometimes recorded, the score will be unreliable. Asking the same questions on every web form and every call, and storing the answers in fixed fields, does more for scoring than any clever model.
Data protection and fairness
Scoring people based on their personal data is a form of profiling under UK GDPR, so transparency obligations apply and your privacy notice should explain it. Where a solely automated decision has a legal or similarly significant effect on someone, extra rules apply. Using a score to decide who your sales team calls first is unlikely to reach that bar in most cases, but take advice if a score could lead to someone being refused a service. Avoid signals that could act as a stand-in for protected characteristics. This is general guidance, not legal advice.
When this matters
Relevant when your team has more enquiries than time and needs to know who to call first. New Era AI sets up scoring and follow-up sequences as part of CRM automation and lead nurturing, and Echo One can capture the job type and urgency on calls so they feed the score.
Questions
Do small businesses need lead scoring?
Only if there are more enquiries than the team can handle promptly. If you can call every lead the same day, focus on speed of response first.
How many leads do you need for AI lead scoring?
There is no fixed number, but predictive models need a good volume of past leads with recorded outcomes to be reliable. With low volumes, start with simple rules.
Should low-scoring leads be ignored?
No. They should get a lighter-touch response, such as a useful automated email and a booking link, so good leads that scored low by mistake are not lost.