B2B Win Rate Benchmarks: The 47% Counts Proposals

Search b2b win rate benchmarks and page one hands you 21%, 29%, 45% and 47%. Google’s AI Overview stacks several of them into a single band and calls a healthy rate 20% to 35%. Read four of the ranking pages and you will find four different numbers presented as the same fact.

None of those figures is fabricated. They are answers to different questions. One counts every opportunity a team created. One counts only the deals that passed qualification. One counts only the deals that got as far as a written proposal. Put a proposal-stage figure next to an opportunity-stage figure and you have manufactured a range that no market ever had.

This page does three things. It names the five denominators in active use and shows which published figure belongs to each. It ledgers every source on the first page of results with its sample, its population and whether it disclosed a denominator at all. Then it gives you the arithmetic to work out your own rate and the test for whether any benchmark you find can legitimately be compared with it.

What a B2B win rate measures

A B2B win rate is the share of sales opportunities that closed won, expressed as a percentage of a defined set of deals. The definition is uncontroversial. Every argument on this topic is about the second half of that sentence.

Notice that the denominator is deliberately wide. It includes the deals that went nowhere, the ones where the buyer stopped replying, and the ones that died in a budget freeze. Those are real outcomes of real selling effort, and a rate that quietly drops them is measuring competitive skill rather than commercial results.

This is the opposite of the discipline most metrics need. For meeting attendance the honest move is to narrow the denominator until it contains only decisions someone actually made. For win rate the honest move is to refuse to narrow it, because every deal you remove was still a deal your team worked. Where the qualification bar sits determines what enters that denominator in the first place, which is why the gate that decides which leads sales formally accepts ends up setting your benchmark before a single deal is worked.

Five things are called “win rate”

Five unrelated metrics share the name win rate, and only one of them counts B2B sales opportunities. This matters more than it sounds, because the search results for the phrase mix all five, and so does anything trained on those results.

The second row is the one to sit with. Bid and proposal management is an established discipline with its own benchmark, measured against proposals submitted, and it reports 45%. Hold that thought until the ledger, because it explains the figure everyone treats as an outlier.

The five denominators in active use

Win rate has five denominators in published use, and moving between them changes the headline further than any difference in selling ability does. This table is the decoder for everything that follows.

Three of these deserve more than a row. The no-decision question is the largest single lever on the page: most CRM defaults file “no decision” under closed lost, which means a lot of teams believe they are reporting decided deals while actually reporting something wider. Cohort maturity is the second: a rate measured before its cohort resolves reads low for purely structural reasons, and that bias grows with the length of the sales cycle you are measuring across.

The third is revenue weighting, which is the only variant that can move in the opposite direction to all the others. Lose two large deals, win twenty small ones, and your count-based win rate rises while the revenue-weighted version falls.

What the published numbers actually say

Restricting the field to pages that publish a figure with any traceable origin leaves twelve publishers and roughly five original datasets. The ledger below records what each one actually disclosed, not what it is quoted as saying.

Twelve publishers, five original datasets, and not one of them publishes a numerator beside a denominator. Every figure on this results page is a percentage with no counts behind it, which means no reader can reconstruct any of them. Two publishers state which denominator they used, and the two state different ones.

What happens to a number on its way to the AI Overview

RAIN Group’s 47% is the cleanest figure in the set and the most badly handled. Trace it forward one hop at a time and you can watch the qualifier fall off.

No step in that chain is dishonest. The population boundary simply falls off at each hop, and by the last one a proposal-stage figure and an opportunity-stage figure are sitting inside the same set of brackets. That is where “21% to 47%” comes from. It is not a market range. It is two metrics printed next to each other.

The corroboration makes it unambiguous. Bid and proposal teams, a separate discipline with no connection to RAIN, report an average RFP win rate of 45% measured against proposals pursued, up from 43% the year before, with enterprise at 47%. Two independent bodies of evidence measuring the same denominator land within two points of each other. RAIN is not an outlier. It is the only page on this results set measuring what the proposal world has been measuring all along.

One study, four sets of bands

The deal-size table that appears on four different sites comes from one study, and no two publishers reproduce it the same way. Optifai publishes it openly, with a sample and a data period, which makes the drift easy to measure.

The Salesmotion discrepancy is innocent: Optifai expanded the study from 847 companies to 939 and the recap froze at the earlier number. Staleness, not error.

Landbase is a different problem. Its four win-rate ranges track Optifai’s closely, but they have been re-bracketed onto ACV boundaries five to ten times larger, and the attribution to Optifai stayed attached. A $60,000 deal reads as 15% to 22% in the study and 18% to 28% in the recap. The “over $1M” band describes a segment the study never measured. The damage carries downstream too, because a team sizing its pipeline coverage target from a published band can be inverting a number that was silently rescaled somewhere upstream.

How the denominator changes the headline

Two teams with identical results can publish win rates of 18% and 47% without either of them being wrong. The cleanest way to see it is to hold the numerator still and change nothing but the set of deals underneath.

Take one quarter’s cohort: 1,000 opportunities created, 120 still open at measurement, 880 resolved. Of those 880, 60 were disqualified before reaching qualification, leaving 820 qualified and resolved. Of the 820: 180 won, 400 lost to a competitor, 240 lost to no decision. Of the 820, 380 received a written proposal.

A 2.6-fold spread, and not one deal behaved differently. This is an IVRIS illustration built from invented counts, so the arithmetic is reproducible and no external population is being represented. The point is the spread, not the numbers.

What the illustration does show is how little of the published disagreement needs explaining. The ladder lands on 21.95%, close to HubSpot’s 21%. It lands on 31.03%, close to Optifai’s 31% median for small deals. It lands on 47.37%, close to RAIN’s 47% and the proposal world’s 45%. A single unchanged cohort reproduces most of the spread that page one presents as a market disagreement.

Where the no-decision deals go

The largest single jump in that table is the fourth rung, and it comes from moving no-decision losses out of the denominator. Membrain works the same arithmetic on a round hundred: 35 won, 30 lost and 35 abandoned reports as 35% when every outcome counts and 54% when only wins and losses do.

UpliftGTM is the only publisher on this results page that states its denominator excludes stale deals, which puts its industry bands on the 54% side of that split while everything around them sits on the 35% side. Nothing labels this in the search results.

Win rate by segment, with the denominator attached

Win rate benchmarks are only usable when the segment and the denominator travel together, so every row below carries both. There is no blended figure in this table and no column you can read across, because no two rows share a population, a denominator and a statistic type at once.

How much each publisher actually discloses

Every page in the set was scored out of six on whether its figure can be audited: sample, population, period, denominator, method, and a reachable primary source. The score measures auditability, not research quality, and a low score is not an accusation of bad faith.

Of the twelve pages ranking for this term, two state which denominator they used. RAIN Group counts proposals; UpliftGTM counts closed deals with stale opportunities removed. Those two definitions cannot produce comparable numbers, and they are the only two definitions anyone published.

The uncomfortable result is that the highest-scoring source on the page is also the one most often flagged as an outlier, and the lowest-scoring source supplies Google’s industry answer. Auditability and citation frequency are running in opposite directions here.

How to calculate a win rate you can defend

To calculate a defensible win rate, pick a cohort that has fully resolved, separate no-decision losses from competitive ones, and publish the counts beside the percentage. The workflow below takes about three quarters of an hour with CRM report access.

Step six is a position rather than a description. Qualified-resolved is the rung that most published figures sit closest to, it survives changes in lead volume, and it does not let a team flatter itself by discarding the deals that stalled. Where the qualification bar sits is then the honest lever, and it is the same bar that governs which conversations count as real meetings earlier in the funnel.

How to check whether a benchmark applies to you

Run any published win rate through these ten checks before comparing it with your own. If tests four through seven cannot be answered from the source, do not compare, and do not average it with anything.

1. Population. B2B sales opportunities, or proposals, trades, matches, or marketing leads?
2. Segment. Which deal sizes, industries and motions are inside the sample?
3. Cohort rule. Grouped by the date opportunities were created, or the date they closed?
4. Cohort maturity. Had the cohort fully resolved when the rate was measured?
5. Denominator. All created, all resolved, qualified resolved, decided only, or proposals issued?
6. No-decision treatment. Are abandoned deals in the denominator, out of it, or unmentioned?
7. Statistic type. Weighted mean, company mean, median, or a raw cohort rate?
8. Unit. Counts of deals, or revenue-weighted?
9. Recency. What period does the data describe, as opposed to when the page was updated?
10. Source mix. Is this original data, or one dataset quoted by several pages?

Two rules follow. Never average a median with a mean, and never treat repeated citations of one dataset as independent confirmation. The second one is not hypothetical here: the deal-size table appearing on four sites is one study, and the “21% to 47%” range is one metric standing next to a different metric.

Test nine catches the quietest failure. HubSpot’s 21% describes 2023, was published in 2024, and is currently being reproduced by pages dated 2026 with no year attached. Nothing about those pages is inaccurate, and a reader still ends up benchmarking against a three-year-old self-report.

Method, sources and revision history

Evidence for this page was collected in August 2026 from public sources only, working from the first page of Google results for the primary keyword plus the sources those pages cite. IVRIS did not run any of the underlying studies, and each organisation retains ownership of its own research and statistics.

Sources were included in the ledger when the publisher disclosed enough to place the figure: a sample, a population, a data period, or a stated metric definition. Pages quoting precise ranges with no traceable original were recorded as exclusions rather than inputs, which is why several widely repeated percentages, including “20% to 40%” and “below 25% is concerning”, do not appear as ledger rows. Where a primary domain would not serve a direct request, the figure was confirmed through search results quoting the same publisher and that method is noted rather than hidden.

Three figures on this page are IVRIS calculations rather than reported findings, and all three are labelled where they appear: the five-rung denominator ladder, which uses invented counts; the disclosure scores, which apply a six-criterion rubric of our own to public pages; and the count of how many ranking pages state a denominator. The Ebsta and Pavilion “19%” that circulates widely is not ledgered here at all, because that figure has already been traced to secondary coverage rather than to the report itself.

Last reviewed: 9 August 2026. Revision history: v1.0, first publication, 9 August 2026. Suggested citation: IVRIS Tech, “B2B Win Rate Benchmarks: What Each Number Counts”, ivristech.com, 2026.

This page is reviewed when any ledgered publisher updates its figures, when a new original dataset appears, or every six months, whichever comes first. Corrections are welcome, particularly from any publisher whose denominator we have recorded as unstated when it is documented somewhere we did not reach.

Frequently Asked Questions