The figure is real. A June 2021 white paper reports that 76.7% of B2B advertisements scored one star out of five on a creative testing scale, and states in prose that a B2B ad is 154 times more likely to score one star than four or five.

It also measures declared emotion in a test rather than any commercial outcome, was applied to B2B using a model the document admits had never been analysed on B2B advertising, and was produced by two companies that both sell something to advertisers who believe it. None of that is disclosed anywhere in the paper.

This page gives the numbers accurately, then everything you need to cite them honestly.

What the paper actually reports

Worth getting the source right, because two documents from the same publisher get conflated constantly.

The right document. A white paper published in June 2021 by a professional network’s B2B research arm, on creativity and profit. Four named authors, all employees of that network.

Not the other one. A separate 2021 report on B2B effectiveness, covering 435 award-winning cases from two industry effectiveness databases between 2010 and 2021, contains no star-rating distribution at all. If someone cites that one for the 77%, they have not opened it.

The distribution, verbatim. 76.7% of ads score 1 star. 16.8% score 2 stars. 6.0% score 3 stars. 0.4% score 4 stars. 0.1% score 5 stars.

The prose version, verbatim. That 77% of B2B creative is 1 star, supporting the contention that creative is in crisis; that powerful creative does exist but is exceedingly rare; and that only 0.5% of 1,700 B2B ads scored 4 to 5 stars, or put another way, a B2B ad is 154 times more likely to score 1 star than 4 to 5.

So the 77% is the paper’s own rounding of 76.7%, not a press distortion. On that narrow point the citation chain is clean, which is rarer than it should be.

The sample, verbatim. A custom analysis of 6 B2B categories with 1,700 ads, drawn from a testing database described as holding more than 40,000 ads across all sectors, predominantly consumer.

What the document does not disclose. The countries. The collection period. The number of human respondents. For a study whose entire output is a percentage derived from human responses, the absence of a respondent count is a real gap.

Published distribution of star ratings across business-to-business advertisements and the undisclosed elements of the studyChart presenting the star rating distribution published in a June two thousand and twenty-one white paper on business-to-business advertising creativity, together with the elements of that study which the document does not disclose. The published distribution states that seventy-six point seven percent of business-to-business advertisements scored one star, sixteen point eight percent scored two stars, six point zero percent scored three stars, zero point four percent scored four stars and zero point one percent scored five stars. The document states in prose that seventy-seven percent of business-to-business creative is one star, supporting the contention that creative is in crisis, that powerful creative does exist but is exceedingly rare, and that only zero point five percent of the one thousand seven hundred advertisements scored four to five stars, meaning a business-to-business advertisement is one hundred and fifty-four times more likely to score one star than four or five stars. The seventy-seven percent figure is therefore the paper’s own rounding of seventy-six point seven percent rather than a distortion introduced by secondary coverage. The declared sample is a custom analysis of six business-to-business categories comprising one thousand seven hundred advertisements, drawn from a testing database described as containing more than forty thousand advertisements across all sectors and predominantly business-to-consumer. The elements not disclosed anywhere in the document are the countries covered, the period over which the advertisements were collected and tested, and the number of human respondents, the last of which is a substantial gap given that the entire output of the study is a percentage derived from human emotional responses.The distribution, as published1,700 B2B ads, six categories, June 2021.1 star76.7%2 stars16.8%3 stars6.0%4 stars0.4%5 stars0.1%The paper’s own framing”154x more likely to score 1 star than 4-5 stars”The citation chain here is clean”77%” is the paper rounding its own 76.7%.What the document does not disclose anywhereThe countries. The collection period. The number of human respondents.
The percentage is the paper's own rounding, which is clean. The respondent count is simply absent. Source : Cashing In On Creativity, The B2B Institute at LinkedIn (2021)

What the star rating measures

This is where the claim and the method part company, and it is not hidden. It is simply not mentioned in the same breath as the percentage.

The protocol. Respondents are shown an advertisement and asked to indicate how they feel about it. They select one of seven basic emotions, or neutral, from a pictorial scale, drawn from a well-known psychological framework of universal emotions.

What that produces. A measure of declared emotional response to an ad viewed out of any purchasing context. No respondent buys anything. No commercial outcome is observed.

The scale. 1.0 is described as Low, 2.0 Modest, 3.0 Good, 4.0 Strong, 5.0 Exceptional.

What the testing company claims for it. That the star rating predicts long-term brand growth based on an ad’s creative quality, calculated by measuring emotional response.

The word doing the work. Predicts. Not measures. The rating is an input to a model, and the model is where growth appears.

Why this matters for a B2B advertiser. A B2B purchase involves a buying group, a procurement process and a months-long evaluation. Whether an individual felt an emotion while viewing an ad in a testing panel is a long way upstream of that, and the distance is not addressed.

The growth numbers are modelled, not observed

The most quoted consequence of the 77% is that most B2B advertising drives no growth. That specific claim comes from a model, and the model has an assumption in it.

The published relationship. 1 star drives 0% market share growth. 2 stars 0.5%. 3 stars 1.0%. 4 stars 2.0%. 5 stars 3.0%.

The assumption attached. Those figures assume 10% excess share of voice in the category. Change that assumption and the growth numbers change with it.

Where the relationship comes from. A general database of more than 40,000 advertisements, overwhelmingly consumer.

The sentence in the document that should be quoted more often. That there was no specific analysis of B2B ads before this study. The star-to-growth relationship was therefore established elsewhere and applied to B2B, not derived from it.

What that makes the headline conclusion. A consumer-derived model, applied to a B2B sample, generating a B2B growth claim that was never validated against B2B commercial results.

What would settle it. Following those 1,700 advertisers and observing what happened to their market share. That study does not exist, and no version of it is published.

The chain from measured emotional response to claimed market share growth in the business-to-business advertising studyDiagram tracing the chain of inference from what is actually measured in the business-to-business advertising effectiveness study to the growth claim ultimately made from it. What is measured is declared emotional response: respondents are shown an advertisement and asked to indicate how they feel about it, selecting one of seven basic emotions or neutral from a pictorial scale derived from a psychological framework of universal emotions, with no respondent purchasing anything and no commercial outcome observed. That response is converted into a star rating on a scale where one point zero is described as Low, two point zero as Modest, three point zero as Good, four point zero as Strong and five point zero as Exceptional. The star rating is then connected to market share growth by a published relationship stating that one star drives zero percent market share growth, two stars zero point five percent, three stars one point zero percent, four stars two point zero percent and five stars three point zero percent, a relationship that explicitly assumes ten percent excess share of voice in the category, so that changing that assumption changes the growth figures accordingly. That relationship was derived from a general database of more than forty thousand advertisements which is overwhelmingly business-to-consumer, and the study document itself states that no specific analysis of business-to-business advertisements existed before it, meaning the star-to-growth relationship was established in other categories and applied to business-to-business rather than derived from business-to-business data. The resulting headline conclusion is therefore a consumer-derived model applied to a business-to-business sample, producing a business-to-business growth claim never validated against business-to-business commercial results. Settling the question would require following the one thousand seven hundred advertisers and observing what happened to their market share, and no such study is published.Measured, then modelledMeasuredRespondents pick one of sevenemotions, or neutral, from apictorial scale.→ConvertedA star rating, 1 to 5.Low, Modest, Good, Strong,Exceptional.→ModelledMarket share growth, froma mostly consumer database,assuming 10% ESOV.The published relationship, with that assumption baked in1 star0% market share growth4 stars2.0%2 stars0.5%5 stars3.0%3 stars1.0%The sentence in the document that deserves more attentionThat there was “no specific analysis of B2B ads” before this study. So the star-to-growthrelationship was built elsewhere and applied to B2B, never derived from it.
The star rating is measured. The growth figure is modelled from a mostly consumer database, assuming 10% excess share of voice. Source : Cashing In On Creativity, and System1 published methodology (2021)

The number moves

A percentage that changes when the sample changes is a description of a sample, not a property of B2B advertising.

January 2021. Trade press reported 1,600 B2B ads shown to a sample described as 6 million people worldwide over the past four years, with 75% scoring one star or less. This is the only version that gives a panel size, and it is not the version in the white paper.

June 2021. The white paper: 76.7%, on 1,700 ads across six categories, with no panel size given.

August 2022. A representative of the same institute, speaking at a conference, described a most recent analysis of over 600 B2B ads in which 71% scored one star.

What that spread tells you. The database is live and the percentage tracks whatever is currently in it. Three numbers, three samples, eighteen months.

How to cite it honestly. As roughly three quarters of tested B2B ads scoring lowest on one company’s emotional response scale, in samples of several hundred to a couple of thousand, between 2021 and 2022. That is defensible. “77% of B2B ads fail” is not the same statement.

And what does not change across all three. The direction. Whichever sample you take, the overwhelming majority score at the bottom. The finding is unstable in its second digit and robust in its shape.

Three published versions of the business-to-business advertising star rating claim across eighteen monthsTable recording the three distinct published versions of the claim that most business-to-business advertising scores lowest on a creative testing scale, across an eighteen-month period, showing that the percentage tracks whichever sample is current rather than describing a fixed property of business-to-business advertising. In January two thousand and twenty-one, trade press reported that of one thousand six hundred business-to-business advertisements shown to a sample described as six million people worldwide over the previous four years, seventy-five percent scored one star or less; this is the only published version that discloses a human panel size, and it is not the version that appears in the white paper. In June two thousand and twenty-one, the white paper reported seventy-six point seven percent on one thousand seven hundred advertisements across six categories, giving no panel size at all. In August two thousand and twenty-two, a representative of the same institute speaking at a conference described a most recent analysis of over six hundred business-to-business advertisements in which seventy-one percent scored one star. The spread across these three versions indicates that the underlying testing database is live and continuously updated, so the reported percentage reflects its current contents. An honest citation would therefore describe roughly three quarters of tested business-to-business advertisements scoring lowest on one company’s emotional response scale, in samples ranging from several hundred to a couple of thousand, between two thousand and twenty-one and two thousand and twenty-two. What does not change across the three versions is the direction: whichever sample is taken, the overwhelming majority score at the bottom of the scale, so the finding is unstable in its second digit and robust in its shape.Same claim, three numbers, eighteen monthsWhen and whereAds testedOne starPanel size given?Jan 2021, trade press1,60075%Yes: “6 million people”Jun 2021, the white paper1,70076.7%NoAug 2022, conference600+71%NoWhat that spread meansThe database is live. The percentage trackswhatever is currently in it.What does not changeThe direction. Whichever sample you take,the overwhelming majority score at the bottom.An honest citation”Roughly three quarters of tested B2B ads score lowest on one company’s emotion scale.”
The database is live, so the percentage describes whatever is currently in it. The direction holds; the second digit does not. Source : Trade press Jan 2021; white paper June 2021; conference report Aug 2022 (2022)

Who produced it, and what they sell

Not a hidden conflict. An undisclosed one, which is different and in some ways worse.

The two parties. A professional network that sells B2B advertising inventory, and a company that sells creative pre-testing. The white paper is a joint production.

What the paper recommends. Investing more in creative that is tested and reworked continuously. That recommendation is the pre-testing company’s product, and the growth it promises is delivered on the network’s inventory.

What the paper discloses about this. Nothing. Searching the full text for funding, sponsorship, disclosure, limitation or caveat language returns no results. There is no statement of interest and no statement of methodological limits anywhere in the document.

How far the pattern extends. The best-known B2B split of the long-and-short budget research, which produces the 46% brand and 54% activation figure for B2B, was commissioned by the same network. One of its executives described the motivation directly: seeing an opportunity to popularise the researchers’ recommendations among their client base by commissioning a B2B cut of the data.

And the 95:5 rule. Its author, an academic at an independent institute, notes on his own page that he wrote it for the same network in 2021, thanking two of the same executives for prompting him. He also writes that the 95% figure is not meant to be a precise rule and that it is a heuristic to get the idea across.

The honest summary of the field. The underlying case databank is genuinely independent, built from award submissions since 1980. Every B2B analysis of it that anyone cites was commissioned by a company selling B2B advertising.

Commissioning and funding relationships behind the widely cited business-to-business advertising effectiveness findingsTable setting out who commissioned or funded each of the widely cited findings about business-to-business advertising effectiveness. The finding that seventy-seven percent of business-to-business creative scores one star was published jointly by a professional network selling business-to-business advertising inventory and a company selling creative pre-testing services, with no disclosure of funding, sponsorship, commercial interest or methodological limitation anywhere in the document, and the recommendation it makes, namely to invest more in continuously tested and reworked creative, is precisely the pre-testing company’s product delivered on the network’s advertising inventory. The finding that business-to-business budgets should split forty-six percent to brand building and fifty-four percent to activation was produced by independent researchers analysing an independent case databank, but the business-to-business analysis itself was commissioned by the same professional network, whose executive described the motivation as seeing an opportunity to popularise the researchers’ recommendations among their client base by commissioning a business-to-business cut of the data. The ninety-five to five rule stating that only five percent of a market is in a buying cycle at any moment was written by an academic at an independent research institute, who notes on his own page that he wrote it for the same network in two thousand and twenty-one at the prompting of two of the same executives, and who states explicitly that the ninety-five percent figure is not meant to be a precise rule but a heuristic to convey the idea. The underlying case databank drawing on award submissions since nineteen eighty is genuinely independent, but no business-to-business analysis of it exists that was not commissioned by a seller of business-to-business advertising.Who paid for each finding you have been quotingFindingProduced or commissioned byDisclosed?“77% score 1 star”Ad inventory seller + test sellerNoNo funding, sponsorship, limitation or caveat language anywhere in the document”B2B is 46% brand, 54% activation”Commissioned by the same networkStated elsewhere”an opportunity to popularise… by commissioning a B2B cut of the data""Only 5% are in-market”Written for the same networkYes, by the authorAnd he adds: “The 95% figure is not meant to be a precise rule.”What is genuinely independentThe case databank itself, built from awardWhat is notEvery B2B cut of it anyone cites was paidsubmissions since 1980.for by a seller of B2B advertising.None of which makes any of it false. All of which belongs in the citation.
The raw case databank is independent. Every B2B cut of it was paid for by a seller of B2B advertising. Source : Source documents and publisher statements (2021)

What is probably true anyway

Being sceptical about the evidence is not the same as disagreeing with the conclusion.

The observation matches what you can see. Open a feed and look at ten B2B ads. Most are indistinguishable, most name a category rather than a customer, and most would work equally well for a competitor with the logo swapped. You did not need a star rating to notice this.

The mechanism is plausible. An ad nobody remembers cannot influence a purchase considered months later. That does not require an emotional response scale to be true.

The advertisers behind the tested ads chose those ads. So the 77% is also a measurement of what B2B marketing departments approve, which is a governance finding as much as a creative one.

Where I would not go. From “most B2B ads score low on an emotion test” to “your ads produce no growth”. The second statement needs the modelled link, and the modelled link was never validated on B2B.

The reasonable posture. Treat the finding as a strong prior that your advertising is more forgettable than you think, and then test that on your own account rather than accepting a percentage as a verdict.

Two checks worth more than the star rating

Both are free, both take an afternoon, and neither requires buying a test.

The logo swap test. Take your ad, replace your logo and name with a competitor’s. If it still works, it is not your ad. This catches the single most common failure in B2B creative, and it catches it in seconds.

The five-second recall test. Show the ad to ten people outside marketing for five seconds. Ask two questions: who was that from, and what were they offering. If most cannot answer both, distinctiveness is your problem, not budget.

Why these are defensible. They test the two things everyone agrees are necessary: that the ad is attributable to you, and that its offer is comprehensible. No model, no scale, no assumption about excess share of voice.

What to measure after that. Cost per meeting held. If the fixed version of the ad does not move that, the diagnosis was wrong and you have lost a week.

And the trap to avoid. Buying a creative testing subscription because a paper published by a creative testing company said your creative is bad. That may still be the right purchase. It should not be that paper that decides it.

Two free diagnostic tests for business-to-business advertising creativeDiagram presenting two diagnostic tests for business-to-business advertising creative that are free, take an afternoon, and require no purchased testing service. The first is the logo swap test: the advertiser takes the advertisement, replaces their own logo and company name with a competitor’s, and observes whether the advertisement still works. If it does, the advertisement is not attributable to the advertiser, which catches the single most common failure in business-to-business creative and catches it within seconds. The second is the five-second recall test: the advertisement is shown to ten people outside the marketing department for five seconds, after which they are asked two questions, namely who the advertisement was from and what was being offered. If most cannot answer both questions, the problem is distinctiveness rather than budget. Both tests are defensible because they examine the two properties that all parties to this debate agree are necessary, that the advertisement is attributable to the advertiser and that its offer is comprehensible, and neither test requires a model, a rating scale or any assumption about excess share of voice. The measurement to apply after making corrections is cost per meeting held: if the corrected version of the advertisement does not move that figure, the diagnosis was wrong and a week has been lost. The trap to avoid is purchasing a creative testing subscription on the strength of a paper published by a creative testing company stating that the advertiser’s creative is poor; that purchase may still be correct, but it should not be that paper that decides it.Two tests that cost nothingThe logo swapReplace your logo and name with acompetitor’s.If it still works, it is not your ad.Takes seconds. Catches the commonest failure.Five-second recallTen people outside marketing, fiveseconds, then two questions.”Who was that from? What were they offering?”If most cannot answer both, it is distinctiveness.Why these are defensibleThey test the two things everyone in this argument agrees are necessary: attributable, and understood.Then judge on cost per meeting heldAnd do not buy a creative testing subscription because a creative testing company published a papersaying your creative is bad. It may be the right purchase. That paper should not be what decides it.
Attributable to you, and comprehensible. No model, no scale, no assumption about share of voice. Source : Method (2026)

Where to go next

You are choosing between video and static. Video or static ads.

You are working on the opening seconds. The ad hook.

Your creative has been running a long time. Creative fatigue metrics.

You are weighing brand against performance spend. Brand vs performance in B2B.

You want the buying mechanics underneath. Media buying explained.

You are deciding how many creatives to run. How many creatives to test.

In short

  • The figure is real and published: 76.7% of 1,700 tested B2B ads scored one star, rounded to 77% by the paper itself.
  • It measures declared emotion in a test. Respondents pick one of seven emotions, or neutral, from a pictorial scale. Nobody buys anything.
  • The growth claim is modelled, from a mostly consumer database of 40,000+ ads, assuming 10% excess share of voice.
  • The paper states no specific analysis of B2B ads existed before it, so the star-to-growth link was built elsewhere and applied to B2B.
  • The number moves with the sample: 75% on 1,600 ads, 76.7% on 1,700, 71% on 600, across eighteen months.
  • The countries, the period and the respondent count are not disclosed, in a study whose entire output is derived from human responses.
  • No conflict of interest is disclosed, though the paper is produced by a seller of B2B ad inventory and a seller of creative testing.
  • Every B2B cut of the independent case databank was commissioned by a seller of B2B advertising, including the 46/54 split and the 95:5 rule.

Run the logo swap test this week, then judge on cost per meeting held. Book a diagnostic, or see how we approach B2B paid acquisition.