Lead Scoring
Lead scoring is a method of assigning numeric points to prospects based on their attributes and behavior to rank how sales-ready each one is.
Key takeaways
- A score is a weighted sum of fit attributes and behavioural signals.
- Fit and behaviour often work better as two axes than one combined total.
- Weights should be derived from what past customers had in common.
- Without decay, tenure outranks intent and long-time subscribers dominate the top tier.
- Always pass the two or three signals that produced a score alongside it.
In depth
A lead score is a weighted sum. Each signal carries a point value, points are added as signals arrive, and thresholds cut the running total into tiers. Two families of signal are usually kept apart: fit attributes that describe who the contact is, and behaviour that describes what they did. Many teams hold these as two axes rather than one total, because a high-fit contact who has done nothing needs a different action from a low-fit contact who reads every email you send.
Weights should be derived from history rather than intuition. Take the contacts who became customers, find the attributes and actions they shared, and set points in proportion to how strongly each one separated them from contacts who never bought. Two mechanisms then pull scores off course. Signals that accumulate forever let a long-time subscriber outrank a real buyer on tenure alone, and correlated signals double-count a single behaviour. Point decay and negative values are the standard corrections.
Scoring earns its keep only where it changes an action: routing, sequence selection, or alert priority. Set the hot threshold at the volume a team can genuinely work through in a day, not at a round number. A scorecard quiz supplies unusually clean input, because the point value sits on the answer option itself and the respondent selects it deliberately. One session yields both fit and intent points without waiting weeks for email opens to accumulate.
A score compresses many facts into one number, and the compression hides the reason. A rep who receives an 82 learns nothing about which criterion is weak, so the tier should always travel with the two or three signals that produced it. Scoring also needs volume: with a handful of deals a year the historical pattern is mostly noise, and a hand-written rule set will beat a fitted model. Contact scores also say nothing about the account around them.
Example in practice
How to measure it
Judge the model by outcome separation between tiers. For each tier, calculate the share of leads that became opportunities and the share that closed. The model works when each tier converts measurably better than the one beneath it. If hot and warm convert at similar rates, either the threshold sits in the wrong place or the weights are too flat to distinguish anything.
Then watch distribution and lag. Track the share of leads landing in each tier month over month; a model that suddenly puts a third of traffic in the top tier has usually met a new channel rather than a better audience. Measure the time between a lead crossing the threshold and the first human contact, because an accurate score with a slow handoff produces the same result as no score.
Common mistakes
The first mistake is scoring everything that can be tracked. Every page view earns a point, so the model quietly rewards browsing volume instead of buying signal, and newsletter readers float to the top tier until reps stop trusting the number. Restrict the model to signals you could defend out loud as evidence of intent, and assign negative points to real disqualifiers such as a free email domain, a competitor address, or a student job title.
The second is treating the threshold as permanent. A team picks sixty as the hot cutoff, a new channel doubles traffic a quarter later, and the sales queue overflows while the number stays where it was. Thresholds are a capacity decision, not a property of the model, so revisit them whenever volume or headcount moves. Recalibrate the weights themselves on a slower cycle, using deals closed since the last review.
Frequently asked questions
What signals are used in lead scoring?
Scoring combines demographic and firmographic fit, such as role and company size, with behavioral intent like content downloads, email engagement, and quiz answers. Negative signals, such as a personal email domain, can subtract points to filter out poor fits.
What is the difference between rules-based and predictive lead scoring?
Rules-based scoring uses weights a marketer defines manually, making it transparent but maintenance-heavy. Predictive scoring uses a model trained on historical conversions to assign weights automatically, which scales better but needs enough data to be reliable.
How do quiz funnels handle lead scoring?
In a scorecard quiz, each possible answer is assigned points, and the system totals them as the respondent progresses. The final score instantly maps the lead to a tier, so prioritization and routing happen the moment the lead is captured.
How many points should each lead scoring signal be worth?
Set points in proportion to how strongly the signal separated past customers from non-customers, then round to a coarse scale. Three or four levels, such as five, ten, twenty and forty, are easier to explain and maintain than fine-grained values. Precision in individual weights matters far less than getting the ranking of signals right.
Where should I set the lead scoring threshold?
Set it at the number of leads your team can actually work each day, then check that the resulting group converts better than the one below it. Threshold is a capacity decision first and a modelling decision second. Starting from rep capacity avoids the common outcome of a queue that grows faster than anyone can call it.
Should lead scores decay over time?
Yes, for behavioural points. Intent fades, so an email opened last year should not carry the weight of one opened yesterday; a common approach is halving behavioural points after a set window of inactivity. Fit points describing company size or role should not decay, since those attributes change through data updates rather than the passage of time.
What are negative lead scoring points used for?
They remove contacts that accumulate engagement without ever being able to buy. Typical deductions cover free email domains, competitor domains, student or job-seeker titles, unsubscribes, and roles outside your buying committee. Negative points are usually more efficient than adding more positive rules, because they clear the top tier of noise instead of adding to it.
How much data do you need for predictive lead scoring?
Enough closed deals for a pattern to be distinguishable from chance, which in practice means hundreds of won and lost records rather than dozens. Below that, a rules-based model built from what your sales team already knows will perform better and stays explainable. You can move to a predictive model once history accumulates.
Should you score contacts or accounts?
Score both if the deal involves a committee. A contact score ranks individuals for outreach, while an account score aggregates activity across everyone at the company, which is what reveals a buying group forming. Relying only on contact scores hides accounts where five people each engaged a little and no single person crossed the line.