Lead Scoring: Ranking Behaviour, or Assessing Quality at Source
Written by LeadScale
Lead scoring is a good idea that quietly disappoints a lot of teams. A model ranks incoming leads so that sales works the best ones first, which is a sensible thing to want. Then the reality sets in. The model scores a newsletter skimmer highly because they opened five emails, and misses a real buyer who filled in one form and went straight to a call. Sales stops trusting the numbers. Marketing re-tunes the weights, again. The instinct is that the model needs more data or a cleverer algorithm. Often the real problem is one step upstream, in the data the model was given to read.
Lead scoring is the practice of assigning a value to each lead to estimate how likely it is to convert, using a mix of behavioural activity and firmographic fit. This article does not argue against it. Scoring is a useful way to prioritise a large volume of leads, and the aim here is not to replace it. What matters is that a model can only rank what it is given, and much of what determines a lead’s quality was already set, or already knowable, at the moment the lead was captured. Assessing some of that quality at capture makes the model that reads it later more accurate.
This article covers what scoring does, why it is a retrospective act, why its accuracy is capped by the state of the data it reads, why scoring a single contact understates a group decision, and how much quality can be checked at the point of capture instead.
What Lead Scoring Actually Does
Start with the plain version. A scoring model looks at what a lead has done and who they are, and turns that into a number that estimates conversion likelihood. Behavioural scoring adds points for actions: an email opened, a pricing page viewed, a demo requested. Firmographic scoring adds or removes points for fit: the right industry, company size, or job title. Predictive or AI scoring does the same job but learns the weights from past conversions rather than having a person set them by hand.
All of these are useful for one thing: putting a large, undifferentiated pile of leads into a rough order so that finite sales attention goes to the leads most likely to be worth it. That is a real benefit, and a team without any scoring usually works leads in the order they arrive, which is worse. Scoring itself is not the problem. Its accuracy suffers when the inputs it reads are weak, and when the whole exercise runs after the moment most of the quality was decided.
Scoring Ranks Behaviour After the Fact
A scoring model does its work once a lead has been around long enough to accumulate activity. By then, several things about the lead are already fixed. Whether the contact gave valid consent to be marketed to was settled at the form. Whether the email address and job title were accurate was true at capture and has been decaying ever since. Whether the company is a good fit was knowable from what the lead submitted on day one. The model infers quality from a trail of clicks, when a good deal of that quality could have been read directly at the start.
This is also where behavioural scoring shows its well-documented weaknesses. Low-intent actions like email opens tend to be over-weighted, so a diligent newsletter reader, or an automated inbox that opens everything, can accumulate enough points to look like a buyer. Many models are blind to recency, scoring an action from a month ago much like an action from this morning. Scores can be gamed, deliberately or accidentally, by anything that generates activity without intent. And firmographic rules misfire, penalising a real buyer who used a generic email domain or trusting a self-reported job title that was aspirational. None of these is fatal on its own, but together they are why the B2B marketing trade publication Demand Gen Report concluded in 2021 that conventional scoring’s ability to predict sales-readiness is, in its words, “pretty unpredictable”. That line is from 2021, but the predictive and AI models sold since as the answer to it have not resolved the underlying data problem it points to, which is why the same complaint is still current.
Why a Lead Scoring Model Is Only as Good as Its Data
The deeper limit is the state of the fields the model reads. B2B contact data does not sit still. The enrichment firm Cleanlist, in a 2026 analysis citing Dun and Bradstreet, puts the decay at roughly 22.5% a year, around 2% a month, and higher in some industries. Cleanlist sells data enrichment, so it has an interest in the alarming version of that number, but the underlying figure is attributed to a neutral data authority and the direction is not in doubt. People change jobs, companies restructure, and email addresses stop working. A record scored this quarter is often no longer the record that was captured.
The point is sharper coming from inside the orthodoxy. Scott Brinker, an executive at HubSpot, which sells scoring tooling, has put it plainly: when a fifth of your records are outdated, your lead scoring, routing and forecasts are “all built on a foundation of lies”. That is an admission against interest from a company with every reason to talk scoring up. The independent CRM publication CX Today makes the same case from the neutral side, noting that scoring and routing “run on stale fields” and that automation, including AI, tends to scale bad data rather than fix it. Its blunt summary is that “AI will not fix bad data. AI will scale it.”
Put together, this is the ceiling on retrospective scoring. A team can re-tune weights indefinitely and adopt the most sophisticated predictive model on the market, and it will still be ranking records whose underlying fields were wrong or stale before the model ever looked at them. It is worth being clear about what re-tuning actually changes. It adjusts how much weight a demo request or a seniority match carries in the final number. It does nothing about whether the seniority field is still accurate, or whether the email address still works. A model tuned to perfection on Monday is tuned against fields that are slightly more wrong by Friday, and more wrong again by the end of the quarter. The fix is better inputs rather than a cleverer model, which makes it a question of data truth rather than of scoring sophistication.
You Are Scoring One Contact for a Group Decision
There is a second structural mismatch. Individual behavioural scoring ranks one person’s activity, but a B2B purchase is rarely one person’s decision. Gartner’s ongoing research on the B2B buying journey finds that around 99% of B2B purchases are set off by an organisational change, and that buying is non-linear, looping across a set of buying jobs rather than moving down a tidy funnel. Gartner sells advisory services and its framing is not disinterested, but the multi-actor, event-driven shape of B2B buying is well established.
That has a direct consequence for scoring. The contact who fills in the form may be a researcher, a champion, a blocker, or a procurement gatekeeper, and their individual click trail says little about where the account as a whole sits. Gartner also finds that buyers are about 1.8 times more likely to complete a high-quality, low-regret deal when they use a supplier’s digital tools alongside a rep, which is a signal about the account’s engagement, not one person’s score. What a captured record tells you about fit and buying-group role is often more durable than the activity trail of whichever individual happened to fill in the form. Scoring the person is not wrong, but treating that score as a read on the account overstates what one contact’s behaviour can carry.
The practical implication is to weight what the form tells you about the account and the person’s role alongside the behavioural score, rather than letting clicks dominate. A senior contact from a target account who submits once is often a stronger signal than a junior researcher from a poor-fit company who opens every email, yet a click-weighted model can rank them the other way round. The signal architecture work covers how declared, implied and inferred signals differ; the point for scoring is that the declared details captured at the form are frequently the most reliable part of the record, and they are available before any behaviour has accumulated at all.
Assessing Quality at the Point of Capture
If a good deal of quality is set or knowable at capture, the constructive move is to assess it there, before the record ever reaches the CRM and long before a model scores it. Three things can be checked at the form. Compliance: whether the consent is valid, which the UK regulator, the ICO, defines as freely given, specific, informed and unambiguous, given by a clear affirmative action, and which is invalid where there are no clear records, a pre-ticked box, or an unnamed controller. Contact truth: whether the email is real and current and the role is accurate. And value: what the submitted details say about fit and likely buying-group role. Checking compliance, truth and value at the point of capture is the Q=CTV idea, and its purpose here is not to replace the score but to make sure the score is read off inputs that have actually been validated rather than stale ones.
That gives a team something to measure. Call it the input-validity rate: the share of the records a model scores whose compliance and contact fields were validated at capture or re-verified since. Put plainly, it is the number of scored records with a current validation flag divided by the total number of scored records. On a base of 40,000 scored records, if only 26,000 carry a validation check within the decay window, the input-validity rate is about 65%, which means roughly a third of the model’s rankings rest on fields that were never confirmed or have since gone stale. The metric has one honest limit, in that it measures input hygiene rather than the model’s predictive lift, so read it alongside the model’s own accuracy rather than as a substitute for it. As a first diagnostic, though, it answers a question most teams cannot currently answer: how much of my scoring is running on data I have actually checked.
A low rate points at a specific fix rather than a vague one. If a third of scored records carry no current validation, the useful response is not to re-weight the model but to add a check at the form and a re-verification step for ageing records, so that the share of validated inputs climbs over time. The score improves because the thing it reads improves, which is a more durable gain than any single round of weight-tuning. The lead quality definitions and the minimum viable lead thresholds are where those validated inputs get their meaning.
Where to Put the Effort in Lead Scoring
Scoring is worth doing, and for prioritising a large volume of leads it is hard to beat. What limits it is not usually the algorithm. It is that the model ranks behaviour after the fact, on fields that were fixed at capture and have been decaying since, for one contact out of a buying group. Re-tuning the weights does not touch any of those three limits. Fixing and measuring the inputs does.
Most of that fixing happens before the record reaches the CRM, at the point of capture, where a validation layer such as the LeadScale Engine checks a record’s consent, contact accuracy and fit on arrival. Validated, current inputs mean the score the model produces later is a read on something real, and the model can do the prioritising job it is good at without inheriting a foundation of stale data.
If you want a concrete next step, measure your input-validity rate. Take the records your model scored last month, count how many carried a current validation check on their key fields, and divide by the total. That number tells you how much of your scoring is running on data you have confirmed, and whether your next move is a change to the model or a change to what happens at the form.
Frequently Asked Questions
Lead scoring is the practice of assigning a value to each lead to estimate how likely it is to convert, using a mix of behavioural activity, such as email opens and page views, and firmographic fit, such as industry, company size and job title. Predictive or AI scoring does the same job but learns the weights from past conversions rather than having a person set them. It is a useful way to prioritise a large volume of leads so that sales works the most promising ones first, but it ranks records after the fact and is only as accurate as the data it reads.
It works well for what it is designed to do, which is to put a large pile of leads into a rough order of priority. It works less well as a precise prediction of who will buy, because it ranks behaviour retrospectively and its inputs decay. Contact data decays at roughly 22.5% a year, by a 2026 industry estimate citing Dun and Bradstreet, so a model often scores fields that are no longer accurate, and behavioural signals like email opens are easy to accumulate without real intent. Scoring is a good prioritisation tool, not a reliable oracle, and its accuracy rises when its inputs are validated.
Behavioural, or rules-based, scoring assigns points that a person decides in advance: so many for a demo request, so many for the right job title. Predictive, or AI, scoring learns the weights automatically from patterns in past conversions rather than having them set by hand. Predictive scoring can find combinations a person would miss, but both are retrospective, ranking a record after activity has accumulated, and both depend entirely on the quality of the underlying data. A more sophisticated model does not rescue stale or unvalidated inputs.
Usually for one of a few reasons. The inputs have decayed, so the model is scoring fields that were accurate at capture but are no longer true. Low-intent actions like email opens are over-weighted, so skimmers and automated inboxes accumulate points. The model is blind to recency, treating an old action like a fresh one. Or it is scoring one contact’s clicks when the purchase is a group decision. Re-tuning the weights rarely fixes these, because the problem is the data and the timing, not the algorithm.
A good deal of it can. Whether a lead gave valid consent is settled at the form. Whether the email and role are accurate is knowable at capture. And what the submitted details say about fit and buying-group role is available on day one. Checking compliance, contact truth and fit at the point of capture, before the record reaches the CRM, gives any downstream model cleaner inputs to score. It does not replace scoring; it makes sure the score is read off validated data rather than stale or unconsented records.
One simple measure is the input-validity rate: the share of the records a model scores whose compliance and contact fields were validated at capture or re-verified since, calculated as scored records with a current validation flag divided by total scored records. If 26,000 of 40,000 scored records carry a current check, the rate is about 65%, meaning a third of the rankings rest on unconfirmed or stale fields. It measures input hygiene rather than predictive accuracy, so read it alongside the model’s own performance, but it answers a question most teams cannot: how much of the scoring is running on data that was actually checked.
