The MQL Isn't Dead. It's Just Been Driving on Four Flat Tyres.
"The MQL is dead" has been a trope for the better part of a decade — Drift published an entire book making the case at least eight years ago, and the clickbait headlines have barely let up since. This year, the Marketing Qualified Lead officially turns 20. Next year, as Dale W. Harrison put it in Episode 1 of B2B Effectiveness: Evidence-Based Marketing Ideas for B2B Practitioners, with our own Liam Moroney, "it won't have to keep using a fake ID to get into the bars."
So why won't it die?
That question was the whole premise of Episode 1 of B2B Effectiveness: Evidence-Based Marketing Ideas for B2B Practitioners, where Liam Moroney sat down with Dale W. Harrison — a data-driven B2B marketing strategist known for his contrarian, evidence-first approach on lead scoring, demand generation, and the metrics the industry loves to misuse. What follows draws heavily on that conversation.
Why the "MQL is dead" myth won't die
Because the underlying job the MQL exists to do isn't going anywhere. The sales team needs leads. That was true a hundred years ago, it's true today, and it'll be true a hundred years from now. If there's no function in the business generating leads for sales to work, sales spends its time on cold outreach instead of closing — one prospect at a time, at a fraction of the efficiency marketing gets from a single campaign reaching thousands.
As long as a company has a department called sales and a department called marketing, there has to be a handoff point — a moment where marketing decides a contact is now worthy of a rep's time. Call it an MQL, an MQA, an MQO, or Bob. It doesn't matter what you call it. The function is structural, not fashionable.
The MQL concept itself is a lot younger than the job it's trying to solve, though. Marketing has been handing leads to sales since long before anyone coined an acronym for it — it's the entire premise of Glengarry Glen Ross. The MQL, specifically, traces back to a paper from a consulting firm roughly 20 years ago, introduced as part of the SiriusDecisions demand waterfall model. Forrester has been flogging the same framework ever since it acquired the firm.
So if the concept isn't dead, why does it feel broken? Because it never really worked in the first place — and almost nobody has been willing to say so out loud.
The real problem was never the M or the L. It was the Q.
Marketing creating leads isn't controversial. What's controversial — and what's actually fallen apart — is what makes a lead qualified.
Harrison tells a story from a RevOps team at a CRM company, roughly seven or eight years ago, that illustrates the problem better than any whitepaper could. Frustrated with their own elaborate, points-based lead scoring model, the team quietly started assigning random scores to a subset of contacts using a random number generator, then mixed those randomly-scored leads in with the ones the "real" model had scored and passed to sales.
There was no difference in the close rate between the two groups.
That's not a scoring problem at the margins. That's a scoring model with zero predictive value, dressed up in enough complexity to look scientific.
The original sin sitting underneath almost every lead scoring model built in the last 20 years is a simple, false equation: engagement data equals intent data. Intent is an unobservable internal mental state — it lives inside someone's head, and until we can safely open people's skulls and look, there is no such thing as intent data. What most vendors sell as "intent" is really just behavioural or engagement data wearing a more expensive label.
The absurdity is easy to demonstrate. In a standard points-based lead scoring system, downloading two ebooks makes you officially twice as likely to buy as someone who downloaded one. Download ten, and you're apparently ten times as likely to buy. Does downloading ten ebooks actually make anyone ten times more ready to purchase? Of course not — but that's precisely what the model is telling your sales team.
Why more data doesn't mean more information
This is where most of the industry's recent "innovation" has gone wrong. AI-driven lead scoring promises to hoover up ten times as many signals. But more data isn't automatically more information.
- Data is a bucket. An empty bucket may or may not contain useful information.
- Ten times as many buckets doesn't mean ten times as much information — it just means more buckets.
- The AI-driven MQL repeats this fallacy at scale, and Marketing Qualified Accounts (MQAs) repeat it again, simply adding up the same flawed points across ten people at one company instead of one. If the underlying logic doesn't work for an individual, it doesn't magically start working when you multiply it by headcount.
Long-term Gartner benchmark data on the B2B sales funnel backs this up starkly. Conversion rates from MQL through to closed-won are consistently low enough to be compatible with a much blunter theory: the best marketing can reliably tell you is whether someone fits the ideal customer profile (ICP) — not whether they're ready to buy right now. Even once a sales opportunity is fully qualified, the benchmark close rate sits at roughly 15–20%. That number holds steady whether a company has 2% market share or 20%, which tells you something important: what changes with market share isn't the odds of winning a deal once you're in the room — it's how often you get into the room at all.
This is the "95-5 rule" in practice: out of a hundred contacts who genuinely fit the ICP, maybe five are in-market right now. The other 95 might buy eventually, from someone, at some point — but there's no reliable way to isolate which five in the moment. Anyone who tells you they can identify exactly who's in that 5% with confidence is, to put it plainly, selling snake oil. If it were solvable, the answer would already exist — the industry has spent trillions of dollars and a quarter of a century trying.
What sales actually needs: thinking in bets, not false certainty
Here's the shift Harrison argues actually matters. An MQL was never really a certainty — it was always a bet. Sales already understands this instinctively: a rep knows that most people they talk to will never buy from them, and that success is a probability game, not a guarantee.
What a scoring system can realistically offer isn't confirmation that a lead is ready to buy today. It's an improved likelihood — a better hand of cards, not a winning one. Three factors, largely absent from twenty years of MQL logic, actually move that probability:
- Are they likely to buy from within the category at all (the ICP-fit question engagement data can genuinely help answer)
- Are they likely to buy from a company that looks like us — a company's market tier and reputation matters enormously here. HubSpot holds roughly 7% overall CRM market share but under half a percent in the Fortune 1000, where Salesforce commands around 85%. Large companies are strongly averse to buying from small vendors, regardless of interest in the category.
- Are we likely to be able to sell to a buyer who looks like them — every sales team, like every athlete, tends to specialise. A team that's built its muscle selling into healthcare and government procurement isn't automatically as strong selling into industrial manufacturing, even with an identical product.
None of these three questions can tell a rep when a deal will close. But together, they can meaningfully shift the odds — and shifting the odds, reliably and repeatedly, is the entire job.
The MQL isn't dead. It just needs new tyres.
Just because a car has a flat tyre doesn't mean it's time to send it to the crusher. You fix the tyre. The MQL is a structural necessity for any business with a sales team and a marketing team sitting on either side of a handoff — what's broken isn't the concept, it's the deterministic scoring logic bolted onto it for the last two decades, built on the false premise that more engagement equals more readiness to buy.
The fix isn't a rebrand to "MQA" or "intent signals" while quietly running the same broken points system underneath a new name. It's replacing deterministic certainty with genuinely probabilistic thinking — building models that acknowledge what the data can and can't tell you, rather than pretending it can tell you everything.
That raises the obvious next question: if engagement and firmographic data can't tell you when someone's ready to buy, what factors actually can move the odds — and which of them can we realistically gather enough information on to act on? Harrison and Moroney are picking that thread up in Episode 2 of the podcast, and we'll cover it here as soon as it lands.
