The belief goes like this: give every lead points for what they do and who they are, add them up, and when the total crosses a threshold, say 70, pass the lead to sales because it is “ready”. Marketing platforms make this easy, and it feels scientific. It is a useful idea. It is also widely misread, and the misreading costs sales teams a great deal of wasted time.
First, what scoring is meant to do
A company with a few thousand leads a month cannot call all of them. Scoring is a way of ordering the queue so that the people most likely to become customers get a human’s attention first. That is a ranking job. Nothing in it promises that a high score means a purchase is coming; it says only that, on past evidence, people who look like this convert more often than people who look like that.
Whether the ranking works is something you can test, which is the part most teams skip. Suppose, illustratively, a firm scores 1,000 leads over a quarter and later checks which became customers. The 60 leads scoring 80 or more produced 9 customers, a 15% rate. The 240 scoring 50 to 79 produced 12, which is 5%. The 700 below 50 produced 7, which is 1%. Total: 28 customers from 1,000 leads. The score separates well; the top band converts fifteen times better than the bottom. That is a model worth keeping.
Now imagine the same table with rates of 4%, 3% and 2.5%. Then the score is hardly better than a coin toss, and the sales team’s feeling that “marketing’s leads are junk” is justified. You would not know which of the two situations you were in without looking.
The problem with one number
Here is the heart of why a high score misleads. A single score adds two quite different things.
Fit is about who they are: industry, company size, role, geography, whether they use a technology your product works with. Fit changes slowly.
Engagement is about what they have done recently: opened emails, visited the pricing page, attended a webinar, asked for a demo. Engagement changes by the day.
Add them together and strange things happen. A postgraduate student researching for a project downloads six whitepapers, visits the site eleven times and registers for two webinars. Engagement points pile up. Their score is 85. They have no budget and no authority. Meanwhile the IT head of a bank in your target segment visits the pricing page once and leaves, scoring 40. Whom would you rather call?
The fix many teams adopt is to keep two scores, or a simple grid: fit rated A, B or C, engagement rated 1, 2 or 3. An A1 goes to a salesperson at once. A C1, an enthusiastic outsider, gets nurtured by email, not by a phone call. An A3, a perfect-fit company that has not engaged, is a good target for outbound effort, since the company is right even though the interest has not appeared yet.
Other ways scoring goes quietly wrong
- Points were guessed once and never checked. Ten points for an email click is an opinion. Compare it with what converted customers actually did before they bought.
- Activity never decays. A person who attended a webinar fourteen months ago still carries those points. Good models let engagement fade over weeks.
- Counting low-value actions. Email opens are increasingly unreliable, since some mail apps load images automatically and register an open that no human made.
- Negative scoring is missing. A personal email address, a competitor’s domain, a job title such as intern or consultant, or a request to unsubscribe should reduce the score.
- Sales never gives feedback. If reps reject leads without recording why, the scoring model has nothing to learn from.
Some behaviours deserve a rule, not a score
A request for a demo, a contact-sales form, or a message saying “please call me about pricing” does not need to climb a ladder of points. These are explicit statements of intent. Route them straight to a person within minutes, whatever the score says, and then check fit. Scoring is for the ambiguous middle, not for the people who have raised their hands.
Making it honest
Review the model each quarter with a table like the one above: score band, number of leads, number that became opportunities, number that became customers. Start from the customers you already have, and look at what they had in common before they bought; build the points from that, not from what sounds plausible. If you have fewer than a few dozen conversions, be cautious about fine-tuning, because small samples produce confident-looking nonsense. A simple rule that sales understands and trusts beats a complicated one nobody checks.
Machine-learning scoring, offered by many platforms, can find patterns a human would miss, but it needs enough historical data to work with and it inherits the biases of your past, such as only ever selling to large firms in Mumbai. It is a tool to test against the simple model, not a replacement for looking at the numbers.
Figures and names in the examples are illustrative.
Leave a Reply