Lead Quality Score
A lead quality score is a numeric value assigned to a lead that summarizes its fit and buying intent, letting teams rank and prioritize prospects rather than treating every lead equally.
Key takeaways
- Weights set the ranking, thresholds set the action; only thresholds change behaviour.
- Carry fit and intent as two axes rather than one combined total.
- Without decay, old browsing outranks a pricing request made this morning.
- Negative points for competitors, students and free email domains keep tiers clean.
- A model nobody recalibrates against closed deals drifts quietly out of usefulness.
In depth
The score is built by attaching point values to attributes and actions, adding them, and cutting the total at thresholds. Most models keep two ledgers: firmographic points for who the person is, and behavioural points for what they did, often reported as separate axes rather than one sum. Points can be negative, so a free email domain or a competitor's company subtracts. Thresholds convert the total into tiers, and it is the thresholds, not the raw number, that decide what happens next.
Weights decide everything, and at the start they are usually set by argument rather than evidence. Making one attribute dominant collapses the score into a proxy for that attribute, while spreading points evenly leaves the total moving too little to separate anyone. Behavioural points need decay, otherwise someone who read three articles last spring outranks someone who requested pricing this morning. A wide point range ranks more finely but demands more data to justify; a coarse three-band model is easier to keep honest.
Teams normally begin with a handful of rules, watch where the tiers land, and adjust until each band holds a workable number of leads per week. Totals are recomputed as new data arrives, so a record moves between tiers over time and routing must tolerate that. In a scorecard funnel the questions carry the points directly: each answer adds to a category and an overall total, the matched tier selects the result page the respondent sees, and that same total travels into the CRM as the routing key.
A score compresses, and compression loses information. Two leads with an identical total can be nothing alike: one a perfect fit with no urgency, the other urgent but far too small to buy. That is why fit and intent are often carried as two numbers instead of added together. Models also inherit the bias of the history that trained them, so a score built on past customers keeps pointing at past customers and misses an emerging segment. Timing stays invisible throughout.
Example in practice
How to measure it
Validate by bucketing recent leads on their score at handover and comparing conversion within each bucket. The curve should climb with the score, and the top bucket should convert several times better than the bottom. Flat stretches mean those points are not earning their place. Check bucket sizes as well, because a tier that holds nearly everyone routes nothing and hides the problem.
Then watch distribution drift: the share of leads landing in each tier, month by month. A sudden swelling of the top tier usually means one rule is firing on a new campaign rather than real demand arriving. Pair that with the rate at which sales rejects top-tier leads, which is the fastest available sign that a threshold needs moving.
Common mistakes
The usual failure is a model nobody ever recalibrates. Point values are argued out in one workshop, entered into the automation tool, and left untouched while the product moves upmarket and the pricing changes. Two years on, the top tier is full of leads sales does not want, and reps quietly ignore the tag. Set a recurring review that tests each rule against won and lost outcomes and deletes the rules predicting nothing.
The other is scoring everything that happens to be trackable. Email opens, generic page views and social clicks earn points because the data exists, not because it predicts anything, and the total gradually measures curiosity instead of buying intent. Keep the model small: a few high-signal rules, each traceable to a real difference in conversion. If deleting a rule leaves the ranking unchanged, that rule was noise with a point value.
Frequently asked questions
How are lead quality scores calculated?
Each attribute or answer is assigned a point weight, and the points sum into a total that maps onto tiers. The weights should reflect which characteristics historically predicted conversion in your data.
Should scores be recalibrated over time?
Yes, because buyer behavior and your ideal customer profile shift, so static weights drift out of date. Review the model quarterly against actual closed-won data and adjust the point values.
What is a good threshold for a hot lead?
There is no universal number; the threshold should be set where conversion rates clearly jump in your historical data. Many teams reserve the top 20 to 30 percent of scores for immediate sales follow-up.
What is a good threshold for a lead quality score?
There is no universal number, because thresholds depend on your point scale and your sales capacity. Set them by working backwards: decide how many leads a rep can genuinely work each week, then place the top threshold where roughly that many records land. Adjust when capacity or volume changes, and check that conversion still differs meaningfully across the resulting bands.
How many points should each attribute be worth?
Assign points in proportion to how strongly the attribute separated won deals from lost ones in your own history. Start coarse, using a few tiers of value such as low, medium and strong, rather than arguing over single points. Precision that the data cannot support only creates false confidence. Refine the weights once you have enough outcomes to compare.
Should a scoring model use negative points?
Yes, for signals that reliably indicate a non-buyer: competitor domains, student or job-seeker roles, countries you cannot serve, and unsubscribes. Negative points stop a genuinely unqualified record from climbing into the top tier through activity alone. Keep them few and specific, because broad negative rules quietly suppress whole segments and are hard to notice once in place.
How often should a lead score be recalibrated?
Review quarterly, and immediately after any change to pricing, packaging or target market. The quarterly pass compares tier conversion against the previous period and retires rules that no longer separate outcomes. A pricing or positioning change invalidates assumptions faster than a calendar does, so treat it as a trigger rather than waiting for the next scheduled review.
Do you need machine learning to score leads?
Not to start. Rule-based scoring is transparent, easy to debug and good enough when a handful of attributes do most of the separating. Models learned from data help once you have thousands of leads with recorded outcomes and too many weak signals to weigh by hand. The limiting factor is usually outcome data quality, not the algorithm.
What is the difference between a lead score and a lead grade?
A grade normally expresses fit as a letter, while a score expresses engagement as a number, and many teams use both together, such as A4 or C1. The split keeps the two dimensions readable: a well-fitting but quiet lead and an active but unsuitable one look different, whereas a single combined total would rank them identically.