What is the average SDR ramp time in 2026?

The Bridge Group's 2025 report, published 6 February 2025 across 351 B2B companies, states 3.0 months, which it describes as the lowest since 2010. The 3.2 months widely in circulation is the figure from its 2023 edition, which surveyed 365 companies. Both are real. The sample skews 78% North America and 83% B2B SaaS, at $47M median revenue, so it is a poor target for an enterprise motion with a much larger deal size.

Is average SDR tenure really 14 months?

Not according to the primary. The Bridge Group's 2025 report states 1.9 years, described as the highest since the early 2010s. The 14-month figure most plausibly comes from its 2016 research, which reported tenure at an all-time low of 1.4 years and roughly fourteen months at full productivity, which is a productivity window rather than tenure. That origin is the best available explanation rather than a proven one, because the 2016 report is gated.

Does calling a lead within 5 minutes really make you 100 times more likely to reach them?

The study says the odds of contact drop 100 times between a 5-minute and a 30-minute callback. That is an odds ratio, not a rate, and it measures a phone connection with a live person rather than a reply. It studied inbound web-form leads at six companies inside InsideSales.com's own software, and states that it did not address close ratios. Applying it to cold outbound is a category error, because a cold prospect never submitted a form.

How common are misquoted statistics?

A 2025 meta-analysis in Research Integrity and Peer Review covering 32,074 quotations across 46 studies found a pooled inaccuracy rate of 16.9%, with about half the errors classed as minor. Two earlier reviews put the range at 14.5% and 25.4%. Those figures come from refereed medical journals, where editorial rigour correlates measurably with accuracy, so they are plausibly a floor for unrefereed commercial content rather than a base rate for it.

How do I check whether a benchmark applies to my company?

Require five fields before the number enters a decision: the edition and publication date, the sample size, the population including geography and segment and revenue band and deal size, the metric's exact definition, and what the publisher sells. Then compare the population to yours. A number missing any of the five is an anecdote with a decimal point.

Why do vendors publish benchmark reports at all?

Because a benchmark creates the anxiety their product resolves, and because gated reports generate leads. That does not make the numbers false. The Bridge Group's methodology is unusually well disclosed. It does predict which way an ambiguity gets resolved, so name what the publisher sells next to the figure and ask who benefits from the direction of any error you find.

Is AI making statistic drift worse?

There is no measurement either way. The largest quotation-accuracy dataset covers quotations up to 2022, before widespread generative summarising, so it cannot answer the question. The failure modes in traced cases, such as pulling a value from an adjacent table cell or attributing to an edition that does not exist, are consistent with automated extraction, but consistency is not evidence.

← Blog

Proof·15 min read

Why Do Published Sales Benchmarks Disagree With Each Other?

The Bridge Group's 2025 report says 3.0 months ramp. Aggregators say 3.2. Both are real numbers, and the publisher seeded some of the confusion itself.

TathagataFounder, ParaphrasePublished September 29, 2026

Last updated: 29 September 2026

You are planning an SDR hire. One page says ramp takes 3.0 months, another says 3.2. One says tenure is 1.9 years, another says 14 months. Both cite the same research firm.

I went and checked, because I had quoted some of these myself. The answer is less flattering to everyone involved than "aggregators are sloppy", and it changed how I read any benchmark.

Here is what the 2025 report actually says, from The Bridge Group's own findings page: 3.0 months average ramp, described as the lowest since 2010. 40% median annual attrition, and the report breaks it down: 13% involuntary, 11% voluntary, 16% promotions. $80,000 median on-target earnings. 60% of reps at quota, the lowest in the study's history. 1.9 years average tenure, the highest since the early 2010s.

Every circulating figure is a real number. None was invented. Each one is detached from its edition, its metric, or its publisher, which is a different failure from being made up and needs a different defence.

The decomposition is worth pausing on, because it is checkable. 13 plus 11 plus 16 is 40. The attrition split is internally coherent, which tells you it was reported rather than reconstructed. It also tells you something about a figure I will come back to: anyone explaining a "34% turnover" number as involuntary plus voluntary is wrong, because that is 24.

And the sample matters more than any single figure. 78% North America. 83% B2B SaaS. $47M median revenue. $50K median deal size. If that is not your company, this is context, not a target.

The publisher's hub page calls it SDR Models, Motions & Metrics: 2025 Research Report (10th ed.). The findings page calls it Sales Development Models, Metrics & Compensation: 2025 Research Report. Same edition. One says the data was collected in 2025; the other says 2024 to 2025. Each page names a different canonical URL for citing it.

The 2007 lead-response study does something similar to itself, which I will come to.

There is a second structural problem. The most linkable Bridge Group URL is a registration gate, and the figures live at a different URL. A writer who lands on the gate and does not fill in the form has nothing to quote and has to take the number from somebody else's page. That is how a chain starts, and no individual in it has done anything dishonest.

Nobody in this chain has to be dishonest for the number to arrive wrong. That is what makes it hard to defend against by judging sources, and easy to defend against by checking units and editions.

I should also correct something I believed going in. I assumed the aggregators were presenting these figures as one study. At least one of them is not. The page most responsible for circulating the drifted numbers describes its own table as directional snapshots from different sources and years, tells readers to verify the originals, and attributes the 57% quota figure to RepVue rather than to The Bridge Group.

The table is honest. My eye merged it. Part of the conflation was mine.

This one is traceable, and you can watch it happen.

Drawn by hand deliberately. The negative half of this is certain: 1.9 years is the 2025 figure and the primary contradicts 14 months directly. The positive half, that it came from 2016, rests on a secondary source and is the best available explanation rather than a proven one.

One aggregator page carries, in adjacent rows of the same table, "14 to 18 months median SDR tenure" and "1.4 year average SDR tenure", both attributed to The Bridge Group and both dated to an edition that does not exist. The decimal slip and its output are sitting next to each other.

I should be honest about the limit here. The 2016 report is gated, so I verified its figures only through a secondary write-up. The negative half of the claim is certain: 14 months is not the 2025 figure, and the primary contradicts it directly. The positive half, that it came from 2016, is the best available explanation rather than a proven one.

And notice the direction. The pages circulating short tenure and high turnover sell outsourced and offshore SDR labour. A shorter tenure figure makes an in-house team look more fragile and more expensive. That is not proof of intent. It is a prior worth holding.

The 2007 InsideSales.com and MIT study is quoted constantly in cold-outbound advice. Its own inclusion rule was leads captured through a web form. Its title asks how much time you have before web-generated leads go cold. There is nothing in it about cold prospecting, because a cold prospect never filled in a form and has no response window to miss.

The sharpest detail is where the error starts. The 2007 document restates its own odds ratio as a rate one page after stating it correctly, so the first corruption in the chain is the primary's own.

Four things the popular version drops. It is an odds ratio, not a conversion rate. The comparison is specifically 5 minutes against 30 minutes, not fast against slow. "Contact" means a dial connecting with a live person for a threshold duration, which each of the six companies defined differently, between two and six minutes. And the study says, as its own standalone sentence, that it did not address close ratios.

The data came entirely from the vendor's own system. InsideSales sold lead-response software. The finding was that you should buy software making you call faster.

What makes this the sharpest example is where the corruption starts. The document restates its own odds ratio as a rate one page after stating it correctly. The primary seeds the error.

Roughly 83% of quotations are accurate at the central estimate. These come from refereed journals where editorial rigour measurably correlates with accuracy, so treat the range as a floor for unrefereed commercial content rather than a base rate for it.

Three independent estimates put the range at roughly 15% to 25%. The heterogeneity is very high, so the interval matters more than any point estimate. One finding is directly useful: journal impact factor correlates with accuracy, meaning editorial rigour measurably reduces error.

That cuts two ways for us. These figures come from refereed medical journals, with editors and referees and reputational stakes. B2B benchmark roundups have none of those. So 16.5% is plausibly a floor for commercial content, not a base rate for it. But the same evidence says most citation is fine, and a post implying that published benchmarks are broadly unreliable would be overclaiming.

There is one more finding worth carrying. A 2009 analysis in the BMJ traced three early distorting papers generating 7,848 supportive citation pathways over a decade. Repetition does not make a claim true, but it does make it feel true, and there is no mechanism that reverses it.

In every case I traced, no figure was invented. Every wrong number was a real number detached from its metric, its year, or its population.

What can be said carefully is narrower. The failure modes in the cases above are pulling a value from an adjacent table cell, collapsing a unit, attributing confidently to an edition that does not exist, and blending sources into one table. Those are the failure modes automated extraction produces.

That is a consistency argument. It is not evidence, and I am not going to dress it up as one. If you find a real measurement of this, it would be genuinely new, because I could not.

This matters more than it looks, because the economics of publishing unverified content already favour volume over checking. Whether machines made that worse is a separate question from whether it is true.

The last field is the one people skip, and it is the cheapest. It never makes a number false. It predicts which way an ambiguity gets resolved, which is a different and more useful thing.

Search the publisher, not the number. Searching for a benchmark returns aggregators, because numbers are what get optimised for. Searching site:bridgegroupinc.com returns the publisher.

Confirm the edition exists. The Bridge Group is biennial: 2019, 2021, 2023, 2025. There is no 2024 edition. A citation to an edition that does not exist is a guaranteed-corrupt source and needs no further checking. That single test disqualifies most of the pages carrying the drifted figures.

Check the unit, which is where most errors live. Odds are not rates. Years are not months. Median is not average. Promotions out of a role are not turnover. Contact is not a reply, a reply is not qualified, and qualified is not closed.

Recompute the arithmetic. It takes thirty seconds and it catches reconstructions. 13 plus 11 plus 16 is 40, so that split is real.

Read the publisher's own limitations paragraph. The Bridge Group states that its data is observational, that the sample skews toward organisations with engaged sales-development leadership, and that subgroup comparisons are directional. That is better practice than most commercial research, and it is the most-dropped sentence in the entire chain.

Then ask whether the population is yours. Setting a 3.0-month ramp target for a $400,000 enterprise motion because a SaaS survey said so is a management error the survey never invited. The same applies to any agency pricing you are quoted, and to the reply-rate benchmarks everyone compares themselves against.

And the rule that makes the rest work: if you cannot resolve a number to a live URL with a sample size and a date, you cannot use it in a decision. Not use it cautiously. Cannot use it.

What is the average SDR ramp time in 2026?: The Bridge Group's 2025 report, published 6 February 2025 across 351 B2B companies, states 3.0 months, which it describes as the lowest since 2010. The 3.2 months widely in circulation is the figure from its 2023 edition, which surveyed 365 companies. Both are real. The sample skews 78% North America and 83% B2B SaaS, at $47M median revenue, so it is a poor target for an enterprise motion with a much larger deal size.
Is average SDR tenure really 14 months?: Not according to the primary. The Bridge Group's 2025 report states 1.9 years, described as the highest since the early 2010s. The 14-month figure most plausibly comes from its 2016 research, which reported tenure at an all-time low of 1.4 years and roughly fourteen months at full productivity, which is a productivity window rather than tenure. That origin is the best available explanation rather than a proven one, because the 2016 report is gated.
Does calling a lead within 5 minutes really make you 100 times more likely to reach them?: The study says the odds of contact drop 100 times between a 5-minute and a 30-minute callback. That is an odds ratio, not a rate, and it measures a phone connection with a live person rather than a reply. It studied inbound web-form leads at six companies inside InsideSales.com's own software, and states that it did not address close ratios. Applying it to cold outbound is a category error, because a cold prospect never submitted a form.
How common are misquoted statistics?: A 2025 meta-analysis in Research Integrity and Peer Review covering 32,074 quotations across 46 studies found a pooled inaccuracy rate of 16.9%, with about half the errors classed as minor. Two earlier reviews put the range at 14.5% and 25.4%. Those figures come from refereed medical journals, where editorial rigour correlates measurably with accuracy, so they are plausibly a floor for unrefereed commercial content rather than a base rate for it.
How do I check whether a benchmark applies to my company?: Require five fields before the number enters a decision: the edition and publication date, the sample size, the population including geography and segment and revenue band and deal size, the metric's exact definition, and what the publisher sells. Then compare the population to yours. A number missing any of the five is an anecdote with a decimal point.
Why do vendors publish benchmark reports at all?: Because a benchmark creates the anxiety their product resolves, and because gated reports generate leads. That does not make the numbers false. The Bridge Group's methodology is unusually well disclosed. It does predict which way an ambiguity gets resolved, so name what the publisher sells next to the figure and ask who benefits from the direction of any error you find.
Is AI making statistic drift worse?: There is no measurement either way. The largest quotation-accuracy dataset covers quotations up to 2022, before widespread generative summarising, so it cannot answer the question. The failure modes in traced cases, such as pulling a value from an adjacent table cell or attributing to an edition that does not exist, are consistent with automated extraction, but consistency is not evidence.

The instinct when two sources disagree is to find the trustworthy one. That instinct is what fails here, because in every case I traced the primary was reachable, free, and clear, and the error still happened.

What actually protects you is smaller and more mechanical. Confirm the edition exists. Check whether the unit is odds or a rate, years or months. Recompute anything that decomposes. Read the limitations paragraph the summary dropped.

One thing I could not establish, and it is the one I most wanted. Nobody has measured whether generative summarising is accelerating this. The dataset that could answer it stops before the question existed. If you see a confident number attached to that claim, it is almost certainly an example of the thing it describes.

The Bridge Group, SDR Models, Motions & Metrics: 2025 Research Report. 10th edition, 351 B2B companies, published 6 February 2025. Revenue Engine LLC, which sells GTM strategy, enablement and fractional sales leadership consulting.
The Bridge Group, sales development metrics findings page. The free page carrying the 2025 figures, under a different title from the gated hub.
The Bridge Group, 2023 SDR Metrics Report. 9th edition, 365 companies, published 1 March 2023. The source of the 3.2-month ramp figure now attributed to 2025.
InsideSales.com and MIT, Lead Response Management Study. 35-page deck presented 16 October 2007. Six companies, 15,000+ leads, 100,000+ call attempts, drawn from InsideSales.com's own system. InsideSales sold lead-response software.
Archived print of the 2008 web version of the same study. Carries the 5-versus-10-minute figures the deck does not.
Oldroyd, McElheran and Elkington, The Short Life of Online Sales Leads. Harvard Business Review, March 2011. A separate artefact routinely merged with the 2007 study. Body paywalled, so its audit figures are unverified here.
Baethge and Jergas, quotation inaccuracy in medicine. Research Integrity and Peer Review, 23 July 2025. 46 studies, 32,074 quotations, pre-registered, two independent raters. Pooled inaccuracy 16.9%, secondary quotation 5.3%. No commercial interest.
dialfyne.com SDR statistics. The aggregator carrying several drifted figures. It labels its own table as directional and mixed-source, and sells outbound calling services.

#benchmarks#evidence#sdr#sourcing#proof