MQL vs SQL: two teams draw one line

Who draws the line between them

If you are trying to settle an argument about which leads count, the thing to know first is that no outside body has settled it. The criteria are yours to set. HubSpot and two other publishers each define the stages that way.

Every published platform definition ends in the same place: a team decides. That is where the argument you are having really sits.

The standard treatment puts the difference at intent to buy. No measurement sits behind that line, so it tells you where to look and not where to put it.

None of those is a bar. They all point at whoever is holding the pen, which in your case is you and your sales team.

What separates the two in practice

Strip out the interest language and four things separate the two. Who decides. What evidence they use. What the lead does next. And what breaks when the bar sits in the wrong place, too low and too high.

The second row is where two companies stop being comparable. A score is something you set. A booked meeting is an event that either happened or did not.

For example, if one company counts scores and another counts meetings, their conversion rates are not the same measurement. Putting them side by side tells you nothing.

What a marketing qualified lead is

An MQL is whatever your scoring model says it is, which sounds like a dodge until you look at what the platforms ship. HubSpot gives you a ladder of eight stages and defines each rung by which team acted, not by what the lead did.

HubSpot ships the ladder and leaves the rungs to you

1. 1Eight stages, and the two in the middle are the ones this page is about. Every definition in the list names an actor or an event, never a threshold.
2. 2The stage sits on the contact record, so it counts people. A B2B deal is usually several people at one company, which is where this model starts to strain.
3. 3HubSpot also notes that its automatic updates "will only move the stage forward." A lead sales rejects does not fall back on its own, so somebody has to move it or your MQL count keeps climbing.

That forward-only rule has a cost you can count. Say you have a rep working forty MQLs in a month who rejects thirty of them. Unless somebody sends those thirty back, your funnel still shows forty leads that reached sales.

So before you measure anything, find out whether your system can move a lead backwards. If it cannot, the number you are about to calculate is counting a queue, not a decision.

What a sales qualified lead is

An SQL is a lead a salesperson has looked at and kept. That sounds simple. But most funnels run two separate steps here, and mixing them is the most common reason two teams argue about one number.

The two steps are acceptance and qualification. A rep accepting a lead means they agree it was worth sending.

A rep qualifying it means they have talked to the buyer and think there is a deal. The demand waterfall that gave B2B this vocabulary put a separate stage between them for exactly that reason.

A sales accepted lead happens within hours of routing, and it tells marketing whether the bar is set right. That is the fastest feedback marketing ever gets.

A sales qualified lead happens after a conversation, and it tells you whether the signals predicted a real need. An opportunity happens when a deal record exists with a value on it.

Skip the first step and rejected leads look exactly like leads nobody has opened yet. Marketing then gets blamed for both.

The handoff itself still earns its place. Gartner reports that B2B buyers are "1.8 times more likely to complete a high-quality deal" when they pair a supplier’s digital tools with a sales rep. Working through those tools alone does worse.

That measures how buyers use tools and a rep together, not how leads get routed. For example, it says nothing about which lead should reach the rep first.

What it does support is keeping a rep in the path at all. That is the reason to draw this line and draw it well.

What a good MQL to SQL rate looks like

Everybody arrives wanting a number to aim at. A table of them is published online, and you should read it. You still cannot use it as a target, and the reason sits inside the same table.

First Page Sage publishes MQL to SQL rates for thirty industries from its own client data gathered between 2019 and 2025. B2B SaaS sits at 13 percent. Legal Services and Real Estate sit at 10 percent, HVAC and Business Insurance at 26 percent.

Half the table sits in the lowest band and B2B SaaS is in it.

Look at what sits above 20 percent: HVAC, insurance, hotels, heavy equipment. What those industries share is not published alongside the table, so check the pattern in your own market before you trust it.

The second argument is stronger and it comes from the same publisher. Their B2B SaaS funnel report, dated 11 June 2025, breaks the same step down by channel.

It reads 26 percent for PPC, 30 for LinkedIn, 39 for webinars, 46 for email and 51 for SEO. The lowest channel is double the industry figure.

Both numbers can be true at once, and the reason is printed on their own page.

The definition that moves the number

1. 1Their SQL requires that the lead "met or booked a meeting with a salesperson." That is an event, and it sits further down the funnel than a rep simply accepting a lead.
2. 2The channel report covers five sources and states an "assumption of a high level of competence" from the team running them. The industry table has no such condition, and it includes traffic the channel table leaves out.
3. 3The headline says 2026 Report. The publication date under it says 3 October 2024, and the data runs to 2025. A year in a benchmark title is a label, not a measurement date.

A rate is not good or bad on its own. Judge it against a definition, a channel mix and a set of companies, and check all three before you borrow anybody’s number. Then make sure yours is computed the same way twice.

1. Fix the denominator

   Count MQLs created in a period, then follow that same group forward. Dividing this month’s SQLs by this month’s MQLs mixes two different cohorts and moves whenever volume moves.

2. Fix the window

   Decide how long an MQL has to become an SQL before it stops counting. Thirty days and ninety days give you different answers from identical data, and a B2B cycle is long enough that the choice matters.

3. Fix the unit

   One company can send you four contacts. Decide whether that is four MQLs or one, and apply the same rule at both ends.

4. Write the three choices next to the number

   Then anybody comparing your rate to a published one can see immediately whether the comparison holds.

Setting your own bar from sales capacity

If a benchmark cannot give you a target, something has to. The most useful answer we found in practice is also the least glamorous: work backwards from how many conversations your reps can hold in a week.

The MQL threshold should be driven by sales capacity and sales feedback. Qualify enough leads that sales is busy, but not so many that they cannot prioritize what matters.

That turns a definitional argument into arithmetic.

1. Count the conversations you can hold

   Say you have three reps who can each work twenty-five new leads a week properly. That is seventy-five a week, or roughly three hundred a month.

2. Sort last quarter’s leads by score

   Take the scoring model you already have and rank every lead it touched, highest first.

3. Cut at three hundred

   Whatever score sits at that line is your threshold. You did not pick the number. Your capacity picked it.

4. Check what the cut throws away

   Look at the deals you closed last quarter and find where they scored. If a third of them sat below the line, your model is ranking on the wrong signals and no threshold will fix that.

5. Revisit it when capacity moves

   A new rep, a quiet quarter or a longer cycle all change the answer. The bar is a dial, not a definition.

Step four is the one people skip, and it is the one that separates a threshold problem from a scoring problem. Those need different fixes, and no benchmark can tell them apart for you.

When this split is the wrong tool

Sometimes sharpening your MQL definition is wasted work, and the case is common. If several people at one company are evaluating you at once, a model that qualifies one contact at a time is counting the wrong unit.

The organisation that put MQL and SQL into common use said this first. Forrester, which owns the SiriusDecisions demand waterfall, published a new version in 2017 built around demand units instead of leads.

The model changed its unit in 2017

1. 1Forrester defines a demand unit as "a buying group that has been organized to address a need the organization is challenged with." The thing you qualify is the group, not the contact.
2. 2The same post says the waterfall debuted in 2006 and was rearchitected in 2012. Explainers that date it to 2002 are working from something other than the publisher.
3. 3Note what did not happen. The stages were not deleted, so this is not an argument for abandoning qualification. It is an argument about what you are qualifying.

A second case belongs to software you can try before you buy. One publisher places the product qualified lead alongside these two for product-led companies, where the signal is what somebody did inside the product.

So if your best leads arrive through a free trial, a form-fill score is reading the wrong file.

Neither case retires the line. Both change what sits on either side of it, and that is worth settling before you spend a quarter tuning a scoring model.

Settling it is most of what gets sold as sales and marketing alignment, and the benefit figures that programme is usually defended with do not survive being traced.

Suppose your definition is settled: buying against it, at your own price per lead, is our B2B PPC service.