Why are my MQLs not converting to SQLs?

Usually because nobody has written down what an MQL is. Published MQL-to-SQL benchmarks run from 13% (First Page Sage, B2B SaaS, both terms defined) to 68% (Ebsta x Pavilion 2025 GTM Benchmarks, 655,000 opportunities, no definition given anywhere in the report). That 5x spread is a definition gap, not a performance gap.

At a glance

• The ratio measures your form threshold, not your lead quality. Move the MQL line and the same month reports 13%, 33% or 68% with no meeting changing hands.
• Define it once: MQL-to-SQL rate = leads in a dated MQL cohort that a salesperson later accepted as an SQL, divided by the whole cohort, read no sooner than 90 days after the cohort closes. Every figure this page computes uses that denominator; the published benchmarks below do not, which is the whole problem.
• The rungs do not chain. First Page Sage publishes 39% lead-to-MQL and 13% MQL-to-SQL for B2B SaaS but defines MQL differently on each page.
• Four buckets, not one rate. Converted, rejected with a reason, touched but undecided, never touched. Three of the four are operations problems no lead-scoring model fixes.

Why are my MQLs not converting to SQLs?

In the order we usually find them when we audit an inbound funnel, there are four causes — and only the last is the one most teams go looking for first.

One: the MQL gate passes everything. If a form fill is an MQL, your MQL-to-SQL rate is your form-fill-to-meeting rate under another name, and it sits in the low teens forever. Two: nobody ever touched the lead — routing rules, territory gaps, a rep on leave, a queue nobody owns. Three: a rep touched it and never dispositioned it: no accept, no reject, no reason, so your denominator fills with leads nobody ever judged. Four: sales genuinely rejected it because marketing and sales are working from different ideal customer profiles.

Causes two and three are operations failures that look exactly like a lead-quality failure on a dashboard. A lead nobody called and a lead sales rejected are the same zero in your numerator, with completely different fixes.

The benchmark you are comparing against probably never defines “MQL”

The Ebsta x Pavilion 2025 GTM Benchmarks Report is a serious dataset: 655,000 opportunities worth $48 billion, 2,000+ CRO and sales leaders surveyed, and, on the PDF’s own summary page, “Analysis of 387 companies”. Its one-page “State of GTM in 2025 Overview” — numbered page 6, the seventh sheet of the PDF — publishes MQL-to-SQL at 68%, up 32% on the prior year. One caveat on that sample before you cite it: the report’s own landing page puts the same study at 349 companies, not 387, and nothing in either place reconciles the two. We use the PDF’s figure throughout.

We read all 34 pages. The string “MQL” appears exactly twice — once in the opening “State of GTM in 2025” commentary, which only points the reader at the chart, and once as the column heading above the 68% itself — and there is no methodology section, no glossary and no written definition of MQL or SQL anywhere in it. The prior-year 2024 B2B Sales Benchmarks report the +32% is measured against publishes no MQL-to-SQL figure at all, so the baseline is not checkable either.

This is not a criticism of the dataset — it is the reason you cannot compare yourself to it. 68% is almost certainly a sales-accepted-lead rate measured inside CRMs where a rep had already agreed to work the record; your low-teens number is almost certainly a form-fill rate. Both are true, and they are not the same measurement.

MQL-to-SQL, MQL-to-opportunity and lead-to-close are different ratios

Four rungs, four denominators. Cohort each by the entry date of its own denominator, not by the date the conversion happened.

Now the trap. Multiply First Page Sage’s own three B2B SaaS rungs — 39% × 13% × 12% — and you get 0.6% of leads reaching closed-won. Do not publish that number. Its lead-to-MQL report defines an MQL as a lead “determined to be in the company’s target market and match one of its buyer personas” — a fit gate. Its MQL-to-SQL report defines one as a contact that has “indicated intent to make a purchase… been determined to be able to afford the product” — an intent-plus-budget gate with no target-market test in it at all. Same publisher, two pages, two different objects called MQL, with not one criterion in common. The rungs do not chain, and neither will yours if the definition is not written down.

Worked example: how the same 520 meetings produce 13%, 33% and 68%

Illustrative inputs, real arithmetic — substitute your own. A US B2B SaaS team takes 4,000 inbound contacts in a month; 520 become sales-accepted, meeting-booked SQLs. That 520 is the only number the business actually feels.

Nothing about the business changed between the first row and the third: same 4,000 contacts, same 520 meetings, same reps, same month. Reporting Ebsta’s 68% here needs a gate that passes just 19% of inbound (765 of 4,000). The trap runs both ways — tighten the gate and the rate climbs from 13% to 68% with no extra meeting booked; widen it to hit a marketing MQL target and the rate collapses with the business unchanged. Agree in advance that the number you steer by is the 520.

A workable MQL definition: the Four-Gate MQL

Since the benchmark will not give you one, here is a definition you can write as a CRM filter this week. A lead is an MQL only when it clears all four gates.

1. Account gate. The account matches an ICP filter you can express as a query: employee band, industry code, region, tech signal. Build it from your last 20 closed-won accounts, not a persona deck.
2. Person gate. A named human whose job title appears on the buying committee of those same 20 won deals.
3. Action gate. A dated action in the last 30 days that cost the prospect something — booked a demo, started a trial, replied to a human. Not a page view, a gated PDF or an email open.
4. Availability gate. No open opportunity on the account, not a current customer, not already worked by another rep.

Then hold it to the Rejection Rule: an MQL definition your sales team never rejects is not a definition, it is a form fill. Our operating thresholds: rejection under 5% of a cohort means the gate is decorative; over 40% means marketing and sales are running different ICPs; between the two, the gate works and your problem is elsewhere. It is the same discipline as writing down which denominator a set rate uses before comparing yourself to anyone (appointment set rate benchmarks), and it sits underneath scoring rather than beside it — the lead qualification framework shows where BANT, MEDDIC and CHAMP go on top of these gates.

The 30-day cohort test you can run on your own CRM this week

This is the test that tells the four causes apart. Two to three hours the first time, nothing beyond your CRM and a spreadsheet.

1. Pick a closed month at least 90 days old. A cohort read too early always looks broken; the tail of an enterprise MQL is long.
2. Export every lead that reached MQL status that month. Columns: lead ID, account, MQL date, SQL date or null, disposition, rejection reason, source, owner.
3. Cohort by MQL date, never by SQL date. Dividing SQLs created in March by MQLs created in March mixes cohorts: it understates the rate on a growing funnel and flatters it on a shrinking one. It is behind most “our rate collapsed” panics we get asked about.
4. Sort the cohort into four buckets and count them: (a) accepted as SQL, (b) rejected by a rep with a reason, (c) touched but never dispositioned, (d) no activity from anyone, ever.
5. Read the answer off the table below and change one thing per measurement window.

What each cohort result means, and what to fix first

The order matters: rebuilding lead scoring while a third of your MQLs are never called is the most expensive way to not solve this. The wider version — a conversion rate is a chain, and the worst-looking link is rarely the one worth fixing — is in where leads actually die between first contact and closed deal.

What running this yourself actually costs

The export and the cohort are cheap: two to three hours, a CRM report, a spreadsheet, and someone senior enough to delete a dashboard that flatters the team. The expensive part is what keeps the number true — a rep making an accept-or-reject decision with a real reason on every MQL inside five business days, forever. That decays first under quota pressure, and the reject-reason field fills with “not a fit”, which tells you nothing.

Above roughly 1,000 MQLs a month, manual disposition stops happening reliably and you are choosing between staffing it and moving first-touch qualification off your closers. That second path is the one LeadsNow runs — 50,769+ AI-booked sales appointments since 2017, on a pay-per-result model where the charge lands on a booked qualified appointment rather than a retainer or a seat — and the qualification and booking layer is described on the AI sales agents for US teams page. Either way, run the cohort test first: a qualification layer built on an undefined MQL just reproduces the same ratio faster.

Frequently asked questions

What is a good MQL to SQL conversion rate for B2B SaaS?

There is no single good number, because the two most-cited sources disagree by 5x for definitional reasons. First Page Sage publishes 13% for B2B SaaS from client data gathered 2019–2025 across 25+ industries, and defines both terms. The Ebsta x Pavilion 2025 GTM Benchmarks Report publishes 68% from 655,000 opportunities across 387 companies (349 on its landing page), and defines neither. Use 13% only if your gate tests intent and budget the way theirs does; if a form fill is an MQL, your comparable is lower again.

Why is my MQL to SQL rate so much lower than the benchmark I read?

Almost always because your denominator is bigger than theirs: a benchmark built on sales-accepted leads has already removed the records a rep would have thrown out, and yours has not. Run the cohort test, and compare only against a figure whose definition you can read.

How long should I wait before measuring an MQL cohort?

At least 90 days after the cohort month closes, and longer if your median sales cycle is over a quarter. Reading a cohort at 30 days measures your follow-up speed, not your qualification.

Is MQL to SQL the same as MQL to opportunity?

No. An SQL is a lead a salesperson has accepted and usually met; an opportunity is a pipeline record with an amount and a close date, normally created after that meeting happens. MQL-to-opportunity is therefore always the smaller number, and quoting one against a benchmark for the other is the most common way teams conclude they have a problem they do not have.

Should I get rid of MQLs entirely?

Only if you replace them with something dated and countable. Teams that drop MQLs for “pipeline sourced” usually just move the undefined gate one rung down. The useful version is counting meetings held and opportunities created per source, cohorted by first-touch date — costed out in our sales pipeline stages breakdown.