Why Sales Doesn't Trust Your Lead Scores (and How to Fix It)

There's a specific, familiar failure mode in B2B marketing: a lead crosses the MQL threshold, gets routed to sales, and nothing happens. Not a rejection, not feedback - just silence. Ask the rep why, and the answer is rarely dramatic: "it didn't look like a real lead." Multiply that across a few hundred handoffs a quarter and you get a sales team that has quietly learned to treat the MQL flag as noise, working leads from their own instinct instead. At that point, the scoring model isn't wrong in some abstract sense - it has failed at its only real job, which is being trusted enough to act on.

The instinct when this happens is to blame the sales team for not following process. It's usually the wrong diagnosis. A scoring model that routes the wrong leads often enough will get ignored by any reasonably good rep, because ignoring it is the rational response to a signal that isn't predictive. Fixing the trust problem means fixing what the model actually measures, not enforcing compliance with a model that doesn't deserve it.

The Root Cause Isn't the Scoring Model, It's What It Measures

Most broken scoring models share the same structural flaw: they measure engagement without measuring fit, and they were calibrated once, early on, and never revisited as the business changed. A lead scoring model is really answering two different questions - how active is this person, and is this person even the right kind of person - and a huge number of implementations only ever build the first half.

Common Ways Lead Scoring Models Break

Treating all engagement equally

A model that awards the same points for opening a newsletter as it does for requesting a demo is measuring attention, not intent. Passive engagement - opens, a single page visit, a webinar registration out of curiosity - is a weak signal on its own. Requesting a demo, asking a pricing question, or returning to a pricing page multiple times in a week is a categorically stronger one. Models that flatten this distinction inflate scores with low-intent activity and dilute the signal that actually matters.

No negative scoring

Positive-only scoring means a lead can accumulate points indefinitely regardless of fit. Without deductions for clear disqualifiers - a personal email domain, a job title with no purchasing influence, a company size well outside your target range, an unsubscribe - a highly active but poorly fitted lead can outscore a well-fitted one who simply engages less. Negative scoring isn't punitive; it's what keeps the score honest about who's actually likely to buy.

Static thresholds that never get revisited

The number that defines "MQL" is often set once, during initial implementation, based on a best guess rather than observed conversion data. As the business adds products, moves upmarket, or shifts its ideal customer profile, that threshold quietly stops matching reality - but almost nobody goes back and re-tunes it once it's live. A threshold that made sense two years ago can be routing leads today that no longer resemble what actually closes.

No feedback loop from sales

This is the most damaging gap, because it's what keeps the first three problems from ever being noticed. Without a mechanism for sales to flag "this MQL was a waste of time" or "this lead I found myself should have scored higher," marketing has no way to see the model drifting away from reality. The model keeps producing the same routing logic indefinitely, sales keeps quietly discounting it, and neither side has the data to fix the disagreement - they just accumulate mutual frustration instead.

A Concrete Example of Drift

Take a SaaS company whose original scoring model awarded five points for any webinar registration and ten points for a demo request, with an MQL threshold set at twenty-five points. Early on, that worked reasonably well. Over two years, the marketing team ran progressively more webinars - some genuinely educational, some closer to thinly veiled product pitches - and webinar volume grew until registrations alone, stacked across a few sessions, could push a lead over threshold without a single high-intent action involved.

Nobody changed the point values; the environment around them changed instead. The result was a steady rise in MQL volume that looked like a win on a dashboard, while the actual close rate on those MQLs quietly fell, because a growing share of them were webinar-collectors who had never shown direct buying intent. Sales noticed the pattern well before marketing did, because they were the ones working the leads - and by the time it surfaced in a calibration conversation, months of routing had already gone to a threshold that no longer meant what it used to. The fix wasn't a new model; it was re-weighting webinar activity down and demo requests and pricing page visits up, then watching the MQL-to-opportunity rate for the next quarter to confirm the correction actually held.

Building a Model Sales Will Actually Act On

Separate fit from behaviour

Score and grade should be two distinct numbers that combine into a matrix, not one blended figure. Behavioural score measures activity and intent; fit grade measures whether this is the kind of account and contact worth pursuing at all. A lead only qualifies as sales-ready when both clear a bar - high engagement from a poor-fit contact and strong fit with no real engagement are both incomplete signals on their own.

Weight high-intent actions heavily, passive ones lightly

Demo requests, pricing page visits, and direct replies to outreach should carry disproportionately more weight than newsletter opens or a single blog visit. The goal is a score where reaching the MQL threshold through a handful of high-intent actions is common, and reaching it purely through accumulated passive activity is rare or impossible.

Add decay

Engagement from eight months ago shouldn't carry the same weight as engagement from last week. Building in score decay over time keeps the number reflecting current interest rather than a lifetime total, which matters especially for longer B2B buying cycles where a contact's situation and urgency can change substantially between one engagement and the next.

Close the loop with a regular calibration meeting

A recurring session - monthly is usually enough - where marketing and sales review a sample of recent MQLs together, sorted into "this was a good lead" and "this was a waste of time," turns anecdotal frustration into a data source. Patterns show up quickly: a specific job title that consistently scores well but never converts, a behaviour that predicts intent better than the model currently credits it for. This is the single highest-leverage habit for keeping a scoring model trustworthy over time, because it catches drift before it hardens into a sales team that's stopped believing the flag altogether.

The Handoff Definition Matters as Much as the Score

A well-tuned score still fails if MQL, SAL (Sales Accepted Lead), and SQL (Sales Qualified Lead) aren't defined the same way by both teams. If marketing calls a lead an MQL the moment it crosses a score threshold, but sales silently expects it to also mean "a named contact, not a generic inbox," the disagreement isn't really about scoring at all - it's about a definition that was never written down and agreed on jointly. These definitions deserve the same quarterly review as the score itself, ideally in the same calibration meeting, so that "sales-ready" means the same specific thing to whoever is looking at the number, on either side of the handoff.

None of this requires exotic tooling - most of it is a modelling discipline and a recurring conversation that a lot of teams simply skip once the initial setup is done. The teams that keep sales trusting the MQL flag are, almost without exception, the ones still having that conversation regularly, long after the scoring model first went live.