The MQL is dead — but not for the reason most people think. Forrester found fewer than 1% of leads ever close. The MQL didn’t die from AI disruption. It died because the MQL-to-SQL handoff was never governed and the qualification threshold quietly dropped to “not spam.” A propensity model won’t fix that.
What actually killed the MQL
“MQL is dead” has become a conference-panel cliché. The diagnosis usually misses the point. RevSure’s State of B2B Attribution 2025 found 92% of B2B marketers admit their pipeline projections lack precision, and only 10.8% have adopted AI-driven predictive models. The model isn’t the bottleneck. The handoff is.
Cross-industry MQL-to-SQL conversion averages around 13%. Teams that pair behavioural scoring with tight ICP fit see 39-40%. That gap isn’t a scoring algorithm problem. It’s a governance problem. Marketing lowered the MQL threshold to hit volume targets. Sales stopped trusting the flag. The handoff became a firewall instead of a bridge.
Consider a mid-market SaaS company whose MQL-to-SQL rate had fallen sharply over three quarters. Marketing was hitting their MQL number — up sharply — but sales had quietly started ignoring anything tagged “marketing qualified” and working their own outbound instead. On closer inspection, most MQLs didn’t match the agreed ICP criteria. Nobody had changed the rules. The threshold had eroded, flag by flag, until “MQL” meant “filled out a form and has an email address.” Marketing celebrated the volume. Sales celebrated by ignoring it.
Why a propensity model won’t save you
Swapping your scoring model for a propensity model fixes the algorithm. It doesn’t fix the handoff. If sales doesn’t trust the signal, they won’t work the lead — regardless of whether the score comes from a rules engine or a neural net. The problem is ownership, not maths. And until you define what a sales-qualified lead actually is — jointly, with sales — every model you build sits on top of a disagreement nobody has resolved.
The fix
Before you change the model, fix the contract. Define what an MQL actually is — jointly, with sales. Set a follow-up SLA. Track MQL-to-SQL by cohort and by rep. Revisit the threshold quarterly. If conversion drops, raise the bar. Don’t lower the volume target. If you can’t get marketing and sales to agree on the definition in one room, the model doesn’t matter.
The MQL isn’t dead because AI arrived. It’s dead because nobody governed the handoff that gave it meaning. Replace the model without fixing that, and you’ve just built a faster way to deliver leads nobody trusts.