How to Improve Your MQL Definition: 10 Finalists Weigh In

To improve MQL to SQL conversion, start with what your MQL definition counts. Most definitions count activity, and activity is not the same as buying intent.

That was the near-unanimous view when Docket asked B2B practitioners one question in its MQL contest: your MQL decides who gets a rep's time, so rewrite it. Entrants had to name what it misses, what theirs would require and how they would prove it predicts better. Three judges from Docket's marketing team scored the shortlist, and the top answer won a Lucid Gravity.

The 10 finalists disagreed on plenty. Three findings held across almost all of them:

• Every finalist would test before switching. All 10 would run the new definition alongside the old MQL and let downstream results decide.
• Nine said the MQL confuses engagement with intent. The tenth went further: buyers misstate what they need, so check what they have actually done.
• Six measured success per hour of rep time. Not by how many MQLs the new definition produced.

Here is what they proposed, where they split, and how to test a new MQL definition in your own funnel.

What's wrong with the current MQL definition?

Most MQL definitions reward activity, so they qualify people who are curious and miss buyers who are ready.

A typical MQL combines ICP fit with a point threshold built from downloads, webinars, page visits and email clicks. It is easy to automate, which is why it stuck. But points record activity, not a purchase taking shape.

• Four ebooks beat a meeting request. One finalist's model gives a form fill 50 points and a meeting request 100, with a 100-point threshold. Four ebook downloads outscore one request to talk to a person.
• A student and an architect score the same. Another described an ebook worth 25 points whether a student downloads it or a security architect about to issue an RFP. Buyers who researched in an AI assistant first show up as direct traffic, with no score at all.
• Committees stay invisible. Six people at one company can each research alone, and none crosses the threshold.

The judges heard the same diagnosis in almost every answer. As Gary Amaral put it while scoring, “the way lots of people do scoring is you’re actually measuring activity, not intent.”

For the basics, see our MQL glossary entry. For why these gaps cost pipeline, see 5 reasons your MQLs are not converting and how MQL point values get invented.

How did the finalists propose to improve the MQL definition?

Three approaches: keep a score but change what feeds it, require evidence that clears gates, or ask the buyer to state the need.

A score adds signals until a lead crosses a line. A gate holds the lead until each required piece of evidence is there.

What if you keep the score but change what feeds it?

• Engagement plus an intent gate. The prize-winning entry made engagement insufficient on its own: a contact qualifies only when an account or contact-level intent signal backs it up, from intent tools the team already pays for. Weights come from 12 to 24 months of MQL to SQL to SAO outcomes, fitted by a model rather than set by judgment. As they put it, "a score that's earned, not assumed."
• Intent and fit, scored separately. Another finalist required a minimum on both, sorted the rep queue by Intent times Fit, and counted three actions in two weeks for far more than the same three over three months.
• Momentum, weighted. A third weighted fit (25%), research depth (25%), commercial intent (30%) and momentum (20%), with time decay so last week's research outweighs last spring's.

What if a lead has to clear gates instead of a threshold?

Five finalists replaced the threshold with lead qualification criteria that must all be met. The names differ; the logic is close.

• Rep-Time MQL. Fit, evidence of a problem and buying motion must all be true, with the threshold set to maximize qualified pipeline per rep hour.
• Evidence-Qualified Opportunity. Fit, intent and momentum, with an "evidence receipt" showing the rep what the buyer said and what is still unknown.
• Rep-worthy moment. Fit, need, timing and readiness, such as asking about price or bringing in a colleague.
• Buyer-verified next step. The company fits, the buyer names a concrete problem and agrees to a next step with an owner and a date. No approved budget required.
• Buying-Group Readiness. Five gates at the account level. The top entry, covered below.

What if the buyer states the need in exchange for something useful?

One finalist stopped inferring need altogether. Their plan turns the best ebooks into diagnostic prompts a buyer runs in their own AI assistant, ending in a summary of the problem, environment, urgency and what has been ruled out. The buyer brings it back for a peer benchmark, and that return is the MQL. "The MQL stops being a score and becomes a trade," as they put it.

Should you trust what buyers say or what they do?

Both fail in different ways, and the finalists split almost evenly.

Behavior tells you someone is researching. A stated need tells you why.

• The case for what they do. One finalist, who books keynote speakers for events, described a client who stated a $250,000 budget. Research showed the client had never spent more than $50,000. A standard MQL would have scored the stated figure. Their fix: compare the stated budget with verified past spend and propose at the verified level. Others on this side lean on research patterns, bursts of activity or external intent data.
• The case for what they say. Others argue behavior cannot tell interest from readiness. Someone can engage with everything you publish and still be years from buying. For one finalist, a stated "not ready" should outweigh any amount of activity: that person is a buyer for later, to be nurtured rather than routed.

The judges pushed on both sides. On the “do” side, Gary pointed out that past spend can mislead: a company that booked small deals last year may have just raised a round and can now afford more. Lauren McHugh added the reverse case: “Companies grow, companies decline.” On the “say” side, Gary noted that a buyer who says “not yet” isn’t the end of the conversation, because part of selling is showing someone why the time is now.

Three finalists named a conversation with the buyer, by a person or an AI agent, as a stronger signal than any passive activity. One put the core problem simply: the MQL "can confuse attention with readiness."

Should you qualify a person or an account?

In larger deals the buyer is a group, so four finalists moved qualification to the account or the buying initiative.

The top entry made the sharpest case: the MQL qualifies a person, but organizations make the purchase. Their replacement, Buying-Group Readiness, is an account state reached through five gates:

1. Fit: the organization belongs in the addressable market.
2. Problem: there is evidence of a relevant business problem.
3. Intent: the organization is evaluating a solution, not just reading.
4. Convergence: several relevant people are involved, or one took a strong direct-intent action.
5. Momentum: the activity is recent and accelerating.

It attaches to the buying initiative, since one large company can run several at once, and every designation needs an evidence trail a seller can read. Their summary: "The MQL measures activity. BGR measures movement."

Three other finalists made a similar move, so five contacts on one project count as one buying motion, not five leads. For keeping the MQL as a supporting metric, see is MQL the right metric for B2B marketing teams?

How do you test whether a new MQL definition improves MQL to SQL conversion?

Run it next to the old definition, measure what each produces per hour of rep time, and count what the new one misses.

This is where the finalists agreed most. Nobody proposed switching overnight.

Why should you run both definitions side by side?

Learning where the two disagree costs almost nothing. The finalists described four designs:

• Back-test first. Score the last 12 to 24 months of closed deals both ways and see which definition better separated wins from losses. It is a quick check, measured in weeks.
• Shadow mode. Score every new lead both ways without changing who gets contacted. Compare three groups: leads both definitions qualify, leads only the old one qualifies, and leads only the new one qualifies.
• Split test. Route half of eligible traffic or accounts each way, with equal sales capacity on both sides. One finalist proposed random assignment by account.
• Holdout with frozen definitions. The top-scored entry calibrates for three months, then locks the definitions for nine, so nobody moves the threshold when results disappoint.

What should you measure besides conversion rate?

MQL to SQL conversion rate is the starting point: SQLs divided by MQLs from the same cohort, measured with a lag so leads have time to convert. But a stricter definition raises conversion just by producing fewer leads.

So six finalists put rep time in the denominator: qualified pipeline per rep hour, opportunities per 100 rep hours, or rep hours per closed deal. One finalist framed it plainly: "Rep capacity becomes a budget." If sales can work 500 leads properly, the aim is the strongest 500, not 1,500 weaker ones. Sales acceptance, time to opportunity, win rate and revenue round out the picture. For more, see metrics B2B revenue teams should track beyond MQLs.

How do you know the new definition isn't just stricter?

Count what it misses. Seven finalists tracked missed buyers or wrongly qualified leads alongside conversion.

• Track total opportunities, not just the rate, so a stricter filter can't look like a win just by passing fewer leads.
• Look for buyers the old model ignored who converted anyway.
• Measure precision and recall: the share of qualified accounts that went on to become opportunities, and the share of all real opportunities the model caught.

How long should the test run?

At least one full sales cycle. The finalists ranged from 60 days of shadow scoring to a 12-month study. Most landed on 90 days to two quarters for the first read, then followed both groups through to closed-won. Conversion shows up fast; win rate and revenue take longer.

How do you get sales to trust a new MQL definition?

Show reps why each lead qualified, log why they reject leads, and agree on response times once the definition has earned them.

• Show the why. Three finalists ruled out black-box scores. A rep should see the evidence and what is unknown, not a number like 87.
• Log rejections. One plan recorded a reason for every lead sales turned down; another reviewed accepted, rejected and missed leads weekly.
• Earn the SLA. One finalist’s team had no service-level agreement on scored MQLs because sales knew the points were guesswork, so only hand-raisers got worked. Their test aimed to earn that SLA, tracking time to first touch as a measure of trust.
• Define it together. Most plans started with marketing, sales and RevOps agreeing on what counts as evidence.

Where does the AQL fit?

The AQL takes the "what they say" side: qualification through conversation, not a score.

An Agent Qualified Lead (AQL) is a buyer who has articulated their needs, demonstrated fit and agreed to a next step. Docket coined the term. An MQL guesses intent. An AQL confirms it.

Docket's website agent gets there in conversation. It answers buyer questions from content you approve, asks your qualifying questions in plain language, books qualified buyers with the right rep, and writes the conversation back to your CRM so the rep starts the call knowing what the buyer said. At Zenity, the website agent recorded 206 conversations in four weeks and booked 10 meetings in its first active period with no BDR at the top of the funnel, with conversations written back to HubSpot.

Like any of the finalists' definitions, it can be tested side by side with your current MQL. For the full criteria, see what is an Agent Qualified Lead? and AI lead qualification.

Frequently asked questions

How do you improve MQL to SQL conversion?

Change what the MQL counts before you tune follow-up. Require evidence of fit and buying intent, not activity alone, then test the new definition side by side with the old one and measure pipeline per rep hour.

What is a good MQL to SQL ratio?

It depends on how your definition is built, so benchmarks rarely transfer. Measure against your own baseline, and make sure a higher rate does not come from fewer total opportunities.

How is an MQL calculated?

Most teams add points for fit and activity, and a contact becomes an MQL past a threshold. MQL to SQL conversion is SQLs divided by MQLs from the same cohort, times 100.

How do you increase sales qualified leads?

Pass better leads, not more of them. Change what qualifies, then check that pipeline per rep hour and total opportunities both rise.

What makes a good MQL definition?

One sales trusts enough to act on quickly: it requires fit, evidence of a problem and recent buying motion, shows the rep why a lead qualified, and has beaten the old definition in a side-by-side test.

How the finalists were judged

Docket ran the contest from September 16 to October 7, 2026. Entrants answered one question, by voice or text, with Docket’s website agent. 900+ entries were screened against the published criteria, and 10 finalists were scored independently by Gary Amaral (VP Marketing), Lauren McHugh (Head of Marketing) and Lee Gladish (Head of Outbound Growth). The highest aggregate score won. Lucid was a promotional partner. Full details are in the official rules.

‍