The two most-cited MQL to SQL benchmarks disagree by a factor of three. One says the B2B average is 13%. The other says it is 40%. Both are real datasets, both are current, and neither is wrong. They are measuring companies that define MQL differently, which is the whole problem with this metric in one sentence.
So before you compare your number to anything, you need to know which dataset you are comparing against and how it was built. This page gives every benchmark with its source, its sample and its date, then says which ones we could not trace at all.
Updated September 2026. Rewritten to lead with the sourced benchmark question this page ranks for. The previous version published four benchmark figures with no attribution, led with two of them in the title and cover, and mixed numbers from datasets that are not comparable. Every number here now carries its origin. Three widely repeated speed and data statistics the previous version also got wrong are corrected below, and the client specifics and unsourced figures have been removed.
The short version
| Question | The 2026 answer |
|---|---|
| What is the B2B average? | It depends on the dataset. First Page Sage puts the cross-industry average at 13%. Optifai's benchmark of 939 companies puts the B2B average at 40%. Pick one and stay with it. |
| What should a B2B SaaS company expect? | 18 to 22% is the range most 2026 agency benchmarks report. Optifai's SaaS figure is 45%. Your own trailing four quarters beats both. |
| What counts as good? | A rate that is stable quarter over quarter with an MQL definition sales has signed. A number that swings 10 points a quarter is a definition problem, not a performance one. |
| Why do the benchmarks disagree? | Because MQL definitions disagree. Two companies in the same vertical can report 13% and 42% and both be accurate about different things. |
| Is the benchmark worth chasing? | Only below Series B. Above it, buying groups and pipeline created are the better unit, and most of the field has moved. |
| What actually moves the number? | The MQL definition first, response time second, contact data third. In that order, because each one caps the next. |
What is a good MQL to SQL conversion rate in 2026?
There is no single number, and anyone who gives you one without naming their dataset is repeating something they read. Here is what the named sources actually say.
| Segment | Rate | Source and what it is built on |
|---|---|---|
| Cross-industry B2B average | 13% | First Page Sage, MQL to SQL Conversion Rate By Industry 2026 report, drawn from their own client data gathered 2019 to 2025 across 25 or more industries, outliers removed |
| B2B average | 40% | Optifai Sales Ops Benchmark, 939 B2B companies, Q2 2025 to Q1 2026 |
| B2B SaaS | 18 to 22% | The range most 2026 agency benchmarks report, including Understory and Growthspree |
| SaaS, Optifai dataset | 45% | Same 939-company benchmark as the 40% line above |
| Top performers | 35 to 40% | Reported by the same 2026 agency benchmarks as the 18 to 22% range, so comparable to that baseline |
| Best in class | 60%+ | Optifai, sitting on the dataset whose average is 40%, not the one whose average is 13% |
| Enterprise SaaS | 30 to 40% | Reported by the 2026 agency benchmarks, on the basis that inbound enterprise leads arrive further along |
| Mid-market SaaS | 18 to 25% | Same, with a 2 to 4 week qualification cycle |
Read down that table and the disagreement is obvious. First Page Sage and Optifai are measuring the same named metric and land three times apart. That is not a rounding difference or a bad year. It is two different populations with two different ideas of what an MQL is.
Never mix a baseline from one dataset with a target from another.
The previous version of this page said the average was 13% and best in class was above 60%. Those two numbers come from datasets whose averages differ threefold, so the gap between them is mostly methodology, not performance. If you take 13% as your baseline, take 26% from the same source as your strong result.
Where does the 13% number actually come from?
It comes from First Page Sage, an SEO agency, and it is an analysis of their own client data gathered between 2019 and 2025. It is widely attributed to Salesforce on LinkedIn and in blog posts. We could not find a Salesforce publication behind it.
That matters in three ways, none of which invalidates the number.
The window is seven years long. A figure spanning 2019 to 2025 averages across the 2021 demand boom and the 2023 correction, so it is a long-run average and not a 2026 reading.
The population is agency clients, reported as roughly 70% B2B agency clients in the coverage of the dataset. Companies that hire a demand agency skew toward inbound-heavy, volume-led motions, which is exactly the motion that produces a low MQL to SQL rate because the MQL bar sits low by design.
And it is cross-industry. First Page Sage's own industry breakdown ranges from around 11% up to 26%, with HVAC and business insurance at the top and legal services and real estate at the bottom. A single B2B SaaS company has no business comparing itself to the blend.
None of that is a criticism of First Page Sage, who publish their methodology. It is a criticism of every page, ours included until today, that quoted the 13% as though it were a law of nature.
What we removed from this page, and why
Four benchmark numbers ran on this page for months with no source attached: 13% cross-industry, 25% SaaS median, 40% top quartile, and above 60% best in class. Two of them were in the title, the description and the cover image. We went looking for the origin of each one.
The 13% traces cleanly, as above. The 60%+ traces to Optifai's best-in-class tier. The 40% is a conflation: it appears in current sources both as the top-performer figure on the 18 to 22% baseline and as Optifai's B2B average, which are not the same claim. The 25% SaaS median traces to nothing we could find. Searching the exact phrasing mostly returns pages restating it, including, until today, this one.
So the 25% median is gone rather than restated, the 40% is labelled with which baseline it belongs to, and the 60%+ carries the dataset it sits on. That is the honest state of the evidence, and it is more useful than a confident number, because it tells you what your own figure can and cannot be compared to.
Which benchmark applies to your company?
Pick by lead source before you pick by industry, because source explains more of the variance than vertical does.
| MQL source | MQL to SQL rate | Source |
|---|---|---|
| SEO and organic search | 51% | First Page Sage funnel benchmarks, June 2025 |
| 46% | Same | |
| Webinar | 39% | Same |
| 30% | Same | |
| Paid search | 26% | Same |
A company that gets 80% of its MQLs from organic search and one that gets 80% from paid will post very different numbers with identical sales teams and identical definitions. If you report a single blended rate, a shift in marketing mix shows up on your dashboard as a conversion problem, and you will spend a quarter fixing the wrong thing.
The pattern underneath those numbers is intent. Someone who searched for the problem and found you is further along than someone who traded an email address for an ebook, and the gap between the top and bottom of that table is roughly the gap between those two states.
Why your number does not match any benchmark
Most likely because you are calculating a same-month ratio while the benchmark is a cohort rate, and the two only agree when your MQL volume is flat.
Same-month conversion divides the SQLs accepted this month by the MQLs created this month. But the SQLs accepted this month mostly came from MQLs created weeks earlier, when the business was smaller. If MQL volume is growing, you are dividing an older, smaller cohort's output by a newer, larger denominator, and the answer comes out low.
Here is the arithmetic. This is our own derivation, not a published benchmark, and it assumes a roughly constant lag and a roughly constant growth rate, which is an approximation rather than a law:
same-month rate ≈ true cohort rate ÷ (1 + g)^(L/30)
g = monthly growth in MQL volume
L = average days from MQL creation to SQL acceptance
Run it on a team whose true cohort rate is 20%:
| MQL volume trend | Lag | Same-month rate you will report | Error |
|---|---|---|---|
| Flat | 21 days | 20.0% | none |
| Growing 5% a month | 21 days | 19.3% | 0.7 points low |
| Growing 10% a month | 21 days | 18.7% | 1.3 points low |
| Growing 20% a month | 21 days | 17.6% | 2.4 points low |
| Growing 10% a month | 45 days | 17.3% | 2.7 points low |
| Shrinking 10% a month | 21 days | 21.5% | 1.5 points high |
The spread between the fast-growing team and the shrinking team is almost four points, and neither team's conversion actually changed. It is the same effect that makes a growing company's churn rate look flattering: a denominator that is moving faster than the numerator it is paired with.
Two consequences worth acting on. If your MQL volume is growing and your reported rate is flat, your real rate is improving. And if you cut lead volume and your rate jumps the following month, wait a quarter before claiming the win, because some of that jump is arithmetic.
Three numbers everyone quotes that are wrong
These three appear in most guides on this topic. The previous version of this page repeated all three, in the form that is wrong.
"Responding in 5 minutes makes you 100x more likely to convert." The 100x is the odds of making contact, not of qualifying the lead. In the 2007 Lead Response Management study analysed by Dr James Oldroyd, then at MIT Sloan, on InsideSales platform data covering six companies and more than 15,000 leads, calling within 5 minutes rather than 30 raised contact odds roughly 100 times and qualification odds roughly 21 times. Twenty-one times is still an enormous number. It is not the same number, and it is nineteen years old, so treat both as directional.
"Respond within an hour and you are 7x ahead of the 24-hour responder." The 7x comparison is against responding one hour later, not 24 hours later. The March 2011 Harvard Business Review article by Oldroyd, McElheran and Elkington, which audited 2,241 US companies, found firms contacting within an hour were nearly 7 times more likely to qualify the lead than firms responding even an hour later, and more than 60 times more likely than those waiting 24 hours or more. Our previous version merged the two comparisons and understated the 24-hour case by roughly an order of magnitude.
"About 30% of B2B contact data is wrong at any given moment." That mixes up a rate with a stock. The most consistently cited figure is HubSpot's database decay simulation, built on MarketingSherpa research, which puts B2B contact database decay at 2.1% per month, compounding to roughly 22.5% a year. ZeroBounce's 2026 email list decay report, drawn from more than 11 billion verified addresses, puts email decay near 23% a year. Those are annual decay rates on a list that is standing still. How wrong your database is right now depends on when you last enriched it, which is a question you can answer directly instead of guessing with someone else's average.
The practical version of all three: speed matters a lot, stale data costs you real conversations, and the exact multipliers in circulation are older and shakier than the confidence around them suggests.
What actually moves the number
Five things, in this order, because each one caps the ones after it. Fixing routing while your MQL definition is broken just gets bad leads to reps faster.
1. Rewrite the MQL definition with sales in the room
The MQL definition should be a contract, not a marketing artifact. It says: if a person hits these criteria, sales works them within X hours, no debate. Without that agreement, reps reject MQLs whenever pipeline gets tight and your rate swings with the season rather than with performance.
A working definition has three layers, and all three are required.
- Fit: company size, industry, geography, tech stack
- Intent: behaviour in the last 14 days, such as a pricing visit or a demo request
- Role: title seniority and function
With only fit and role you are scoring a list. With only intent you are scoring web traffic. The ICP work this rests on has to exist first, and if it does not, start there instead.
This is also the fix that explains the benchmark spread. A company whose MQL is "filled out any form" and a company whose MQL is "director or above at a 200 to 2,000 person software company who viewed pricing twice this week" will report wildly different rates while running identical sales motions. Neither number is wrong. They are not the same metric.
2. Cut response time, and measure the median rather than the mean
Faster is better, and the honest version of the evidence is in the section above: roughly 21 times the qualification odds at 5 minutes versus 30 in the 2007 study, roughly 7 times within the hour versus the next hour in the 2011 HBR audit. Optifai's current 939-company benchmark reports responses under 5 minutes converting at about twice the rate, which is a far more modest multiplier than the folklore and is drawn from this decade.
Measure the median, not the average. One lead answered in four days drags a mean past the point of usefulness while the median tells you what a typical lead experiences.
The fix is routing rather than headcount. The lead lands in a channel a rep actually watches, with the enrichment already attached and a booking link in the message. Our lead routing rules guide covers the mechanics, and the automation side of it is most of what we build in n8n for revenue teams.
3. Fix the contact data before sales touches it
Enrich and verify every MQL before assignment, because a lead nobody could reach gets marked "no response" and becomes evidence that marketing leads are bad. On the decay numbers above, a list untouched for a year is roughly a fifth wrong, and the errors concentrate in exactly the people you want, since job changes break the title, the email and the direct dial at once.
Waterfall enrichment is the standard pattern: run providers in sequence, stop at the first verified hit, pay once. It is a build-once workflow. The waterfall pattern is the detail, and running it properly inside Clay is most of what our Clay work consists of.
4. Score on recency, not just on fit
Most scoring models are demographics plus form fills, which is a 2012 design that ignores when something happened. A VP at a 500-person company who filled in a form nine months ago is not a hot lead. A director at a 100-person company who hit pricing four times this week is.
Breadcrumbs reports behavioural scoring lifting MQL to SQL conversion by up to 40% against demographic-only models, and the 2026 agency benchmarks put teams running behavioural scoring with a tight ICP at 39 to 40% absolute. Note those are two different kinds of claim, a relative lift and an absolute rate, and they get quoted interchangeably.
A four-tier model weighted on recency and frequency behind a hard ICP gate will beat a fifty-criteria scoring sheet in most portals, because the fifty-criteria sheet is usually unmaintained. Our lead scoring model post has the version we deploy.
5. Close the loop with reason codes
Every rejected MQL gets a reason code: wrong title, wrong company size, no budget, bad timing, out of geography, no real intent. Five or six codes, not twenty. Without them marketing has no signal to tune the definition with and the cycle repeats every quarter.
Review the distribution monthly with both teams in the room. If a quarter of last month's rejections were "wrong title", tighten the title filter. If a third were "no real intent", the behavioural threshold is too loose. This is the step everyone skips because it feels like process for its own sake, and it is the only one that compounds.
How do you measure it without lying to yourself?
By cohort. Take every MQL created in March, track them for 60 days, count how many were accepted as SQLs in that window. That is March's rate. It lags by two months, which is annoying and correct.
Run the same-month number alongside it as a directional read, and use the formula above to predict how far apart they should be. If they diverge by more than the formula says, the gap is a definition or data problem rather than calendar drift.
Then split it, because a blended rate hides everything that matters:
- By source, for the channel-mix effect in the table above
- By rep or SDR, which separates a qualification problem from a coverage problem
- By time from MQL creation to first touch, which is the variable you control fastest
- By rejection reason code, sorted by frequency
The last line is the one that matters. Your trailing four quarters is a better benchmark than anything on this page, because it is the only dataset that shares your MQL definition.
Is the MQL dead in 2026?
For enterprise B2B, largely yes, and the shift is no longer controversial. For companies under roughly $50M ARR it is still the right unit of work.
The case against is structural rather than fashionable. Buyers complete most of their evaluation before contacting a vendor, so a form fill is a trailing indicator of a decision already in progress. And the decision is not one person's: Gartner's current guidance puts a typical buying group for a complex B2B solution at six to ten decision makers, each arriving with four or five pieces of research they gathered alone, against the 5.4 stakeholders Gartner reported in 2015. A single-contact MQL captures one member of that group and tells you nothing about the other five to nine. Forrester's Simon Daniels has put the share of leads that ever reach a closed deal at under 1%.
The widely quoted 6.8 decision makers figure is a 2017 snapshot and is now dated on the low side. If you are using it in a deck, use the current range instead.
What replaces it, in the teams that have moved, is buying group or account scoring built on intent signals rather than form fills, with SQLs, pipeline created and CAC payback as the reported numbers. If you are at $50M ARR with a running ABM motion, the answer to "should we improve MQL to SQL conversion" is probably "stop reporting MQLs and move to opportunity-based reporting."
Most teams reading this are not there. Below Series B the MQL is still the cleanest available signal of whether marketing is feeding qualified work to sales and whether sales is taking it seriously, which makes it the cleanest single chart of sales and marketing alignment you can put in front of both teams. Use it while it is useful, and know what will replace it.
If you need an external anchor, take 18 to 22% for B2B SaaS from the 2026 agency benchmarks and treat 35 to 40% as a strong result on that same baseline. Do not pair it with a best-in-class figure from a dataset whose average is 40%.
Where to start if you are starting from scratch
Four weeks, in this order, because each step produces the data the next one needs.
Week one: audit the last 60 days of MQLs against your written definition. If more than a quarter fail it, the input is broken and nothing downstream will help. Tighten the criteria and rerun in 30 days.
Week two: pull the median time from MQL creation to first sales touch. If it is over four hours, fix routing before anything else. It is the cheapest change available because it needs no new headcount.
Week three: sample 100 MQLs from last quarter that did not convert and check deliverability. If more than a fifth bounced, you have a data problem wearing a conversion problem's clothes.
Week four: ask sales to code the last 50 rejections. The largest bucket is what you change in the definition next quarter.
Want a second pair of eyes on your MQL handoff?
We run a 30-minute review of your funnel definitions, response times and data quality, then tell you the three changes most likely to move your MQL to SQL conversion in 60 days.
Book a review →FAQ
What is a good MQL to SQL conversion rate for B2B SaaS in 2026?
Most 2026 agency benchmarks put B2B SaaS at 18 to 22%, with top performers at 35 to 40%. Optifai's 939-company benchmark, which uses a different population, puts SaaS at 45% against a B2B average of 40%. The cross-industry 13% figure everyone quotes is First Page Sage's analysis of their own client data from 2019 to 2025 and is not a SaaS number. Pick one dataset, note which, and compare against it consistently.
Why do MQL to SQL benchmarks disagree so much?
Because MQL is a definition each company writes for itself. Two companies in the same vertical can report 13% and 42% and both be accurate, because one counts every form fill and the other counts only senior-title, in-ICP contacts showing recent buying behaviour. Population differences explain most of the threefold gap between the major published datasets.
How long should MQL to SQL conversion take?
Under 7 days for inbound and under 21 days for cold-sourced leads is a reasonable working target. If yours is longer, the cause is usually first contact, not the sales process: check the median time to first touch before you look at anything else.
Should marketing or sales own the MQL to SQL number?
Both, jointly. Marketing alone will hit the number by lowering the MQL bar. Sales alone will hit it by rejecting more leads. It belongs on a shared RevOps dashboard with a written definition both teams signed and a monthly review with both in the room.
Is the MQL dead in 2026?
For enterprise B2B running account-based motions, effectively yes, and buying group scoring has replaced it. For companies under roughly $50M ARR it is still the right unit of work for marketing and the right handoff signal for sales. The reason the debate is louder in 2026 is that buying groups have grown to six to ten people on complex deals, and a single-contact MQL describes one of them.
How does AI change MQL to SQL conversion?
Mostly through speed and data quality rather than through better judgement. Enrichment and scoring that used to take hours now take seconds, which closes the response-time gap that the 2007 and 2011 studies both identified as the largest controllable variable. It also makes behavioural scoring cheap enough for a team without a data scientist to run. What it does not fix is a bad MQL definition, which is still the first constraint. Our CRM and RevOps work and AI automation work both start there for that reason.
Sources
None of the sources below was directly reachable from the environment this page was written in. Figures are credited to the organisations that published them and the linked pages are where to verify the current position. Every number in this page should be checkable against one of these, and if you find one that is not, it should not be here.
- Cross-industry and by-industry MQL to SQL rates, and the 2019 to 2025 client-data methodology behind the 13% figure, from First Page Sage's MQL to SQL Conversion Rate By Industry 2026 report
- Conversion by lead source (SEO 51%, email 46%, webinar 39%, LinkedIn 30%, paid search 26%), from First Page Sage's lead-to-MQL funnel benchmarks, June 2025
- The 40% B2B average, 45% SaaS figure, 60%+ best-in-class tier and the sub-5-minute response multiplier, from the Optifai Sales Ops Benchmark of 939 B2B companies, Q2 2025 to Q1 2026
- The 18 to 22% B2B SaaS range, the 35 to 40% top-performer figure and the segment splits, from 2026 benchmark write-ups including Understory and Growthspree
- The 100x contact and 21x qualification multipliers at 5 minutes versus 30, from the 2007 Lead Response Management study of InsideSales platform data analysed by Dr James Oldroyd, then at MIT Sloan, covering six companies and more than 15,000 leads, as documented in Expertise AI's verified speed-to-lead statistics
- The 7x and 60x qualification figures and the 2,241-company audit, from The Short Life of Online Sales Leads, Oldroyd, McElheran and Elkington, Harvard Business Review, March 2011
- B2B contact data decay at 2.1% per month and roughly 22.5% per year, from HubSpot's database decay simulation built on MarketingSherpa research, and the 23% annual email decay figure from ZeroBounce's 2026 email list decay report covering more than 11 billion verified addresses, both as summarised in Cleanlist's 2026 data decay write-up
- Behavioural scoring lifting conversion by up to 40% against demographic-only models, from Breadcrumbs
- Buying group size of six to ten decision makers on complex B2B solutions, each carrying four or five pieces of independent research, from Gartner's B2B buying journey research. The 5.4-stakeholder comparison is Gartner's 2015 reading, and the widely quoted 6.8 figure is a 2017 snapshot
- The under-1% lead-to-closed-deal figure attributed to Forrester principal analyst Simon Daniels, and the buying-group scoring shift, from RevSure's 2026 write-up
- The same-month versus cohort formula and the distortion table are our own arithmetic on stated assumptions, not a published benchmark