The claim that emotional advertising outperforms rational advertising rests on a multiple-choice question, and it is worth reading the wording.

Here it is, from the questionnaire behind the most cited effectiveness database in the field:

“The following question is intended to explore how you think your campaign worked. In your opinion which of the following standard models (if any) of marketing communications effects best describes the way in which the campaign proved effective: Simply because of the information it provided Simply because of the emotions or feelings it touched/how likable it was Salience. Got the brand talked about/made it famous It reinforced existing behaviour rather than changed behaviour (the weak theory) Persuasion. Initially gained interest with information and then added emotional appeal A more complex combination of these or other factors”

That is the classification. There is no independent coding, no content analysis of the creative work, no second coder and no measure of agreement between coders. The label “emotional” is applied retrospectively by whoever wrote the case study, which for these entries is the agency and its client, after the campaign has already been judged good enough to enter into an awards competition.

Now the other side of the correlation, from the same questionnaire:

“Please assess the scale of changes to each of the following measures over the course of the evaluation period (please use your own subjective judgment based on your experience of advertising in the market)”

Three response options: very large, large or substantial, small or negligible.

So both variables in the relationship “emotional advertising works better” are supplied by the same interested party, on the same form, from memory. That does not make the finding false. It makes it a survey of practitioner belief rather than a measurement of advertising.

The two self-declared questionnaire items that generate the emotional versus rational effectiveness findingThe two self declared questionnaire items that generate the finding that emotional advertising outperforms rational advertising, taken from the documentary appendix to the two thousand seven report that established the dataset. The classification item asks the case author, in their opinion, which of six standard models of marketing communications effects best describes the way in which the campaign proved effective, offering the options that it worked simply because of the information it provided, simply because of the emotions or feelings it touched and how likable it was, through salience by getting the brand talked about and making it famous, by reinforcing existing behaviour rather than changing behaviour, through persuasion by initially gaining interest with information and then adding emotional appeal, or through a more complex combination of these or other factors. There is no independent coding of the creative work, no content analysis, no second coder and no measure of inter rater agreement, so the label emotional is applied retrospectively by whoever wrote the case study, which for these entries is the agency and its client, after the campaign has already been judged good enough to enter an awards competition. The outcome item instructs respondents to assess the scale of changes to each of several measures over the evaluation period, explicitly telling them to use their own subjective judgment based on their experience of advertising in the market, and offers three response options of very large, large or substantial, and small or negligible, across sales gain, market share gain, reduction of price sensitivity, customer retention and loyalty, new customer acquisition and penetration, and profit gain. Both sides of the relationship are therefore supplied by the same interested party on the same form from memory, which does not make the finding false but does make it a survey of practitioner belief rather than a measurement of advertising. Response rates on these items, from a base of eight hundred and eighty national cases, were four hundred and ten responses on the classification item and three hundred and fifty nine on the role of emotions item.Both variables come from the same formThe classification”In your opinion which of the followingstandard models best describes the wayin which the campaign proved effective”No independent coding. No content analysis.No second coder. No agreement measure.The result”please use your own subjective judgmentbased on your experience of advertisingin the market”Three options: very large, large orsubstantial, small or negligible.Both filled in by the agency that entered the case, after it was judged good enough to enter an awardscompetition. A survey of practitioner belief, not a measurement of advertising.
One classifies the creative, the other grades the result. Both are filled in by the agency that entered the case. Source : Documentary appendix to Binet and Field, Marketing in the Era of Accountability, IPA and WARC, 2007, questionnaire items 21 and 28 (2007)

The numbers themselves, read carefully

The 2007 table that started this is published, and it is smaller and less decisive than its reputation.

Across 371 cases, the share reporting very large effects came out as follows: information 61 percent, emotions 68 percent, fame 72 percent, persuasion 61 percent, and a combined category at 68 percent.

Three things about that table.

The information cell rests on 33 cases. Out of 880 in the database and 371 answering the question.

The gap is not marked as significant. The document defines its own notation for significance at 99, 95 and 90 percent confidence, and applies it to exactly one row, the reinforcement category, which is marked as significantly worse. The emotions-versus-information difference carries no mark.

And on the companion table, the ranking reverses. Measured against awards won rather than self-reported effects, information scores 74 percent and emotions 73 percent.

Later reports state the conclusion more forcefully while publishing less. One free deck asserts that “Emotional advertising is twice as efficient as rational, and delivers twice the profit”. A 2020 one-pager on the subject carries a chart titled “Emotional advertising is more effective, especially over the long-term” whose vertical axis has no values on it at all. Neither publishes a classification criterion.

Published figures for very large effects by declared influence model, and the same by awards wonPublished figures for the share of cases reporting very large effects, broken down by the declared influence model, and the same breakdown measured against awards won. On the very large effects measure, across a base of three hundred and seventy one cases out of eight hundred and eighty in the database, the information category recorded sixty one percent on thirty three cases, the emotions category recorded sixty eight percent on eighty eight cases, the fame category recorded seventy two percent on thirty six cases, the reinforcement category recorded thirty three percent on nine cases and is the only row in the table marked as statistically significant under the document’s own notation, the persuasion category recorded sixty one percent on one hundred and five cases, and a more complex combination category recorded sixty eight percent on one hundred cases, with a weighted average of sixty five percent. Three observations follow. The information cell rests on only thirty three cases. The gap between emotions at sixty eight percent and information at sixty one percent carries no significance mark, even though the document defines a notation for significance at ninety nine, ninety five and ninety percent confidence and applies it to the reinforcement row. And on the companion table measuring the same categories against awards won rather than against self reported effects, across a base of four hundred and four cases, information scores seventy four percent while emotions scores seventy three percent, so the ordering reverses. Later reports state the conclusion more forcefully while publishing less supporting detail, with one freely available presentation asserting that emotional advertising is twice as efficient as rational advertising and delivers twice the profit, and a two thousand twenty single page document carrying a chart titled emotional advertising is more effective especially over the long term whose vertical axis carries no values at all. Neither publishes a classification criterion for how a campaign is judged emotional or rational.Share reporting very large effectsBase 371 cases of 880 in the databaseFame72%n = 36Emotions68%n = 88More complex68%n = 100Information61%n = 33Persuasion61%n = 105Two things the table shows and the summaries do notOnly one row in the table carries a significance mark, and it is not this comparison.On the companion table measuring awards won: information 74 percent, emotions 73 percent.
Seven points apart, on 33 cases in the smaller cell, and not marked significant. On awards, the order reverses. Source : Documentary appendix to Binet and Field, Marketing in the Era of Accountability, IPA and WARC, 2007, tables S15a and S15b (2007)

The business to business version

The B2B claim comes from a 2019 report, and its authors are commendably direct about what it rests on:

“There are still relatively few B2B cases in The Databank, so sample sizes are small, at less than 50 cases. And those cases may not be typical of B2B marketing in general, because the IPA Databank is biased towards effective campaigns: all were cases entered into the IPA Effectiveness Awards competition… Geographically, they are skewed towards the UK… They also tend to have relatively big budgets.”

And on how the classification is made in that report:

“Case study authors (in the case of B2B cases these will be B2B agencies and their clients) assess how important rational and emotional consideration are in purchase decisions.”

Fewer than fifty cases, split across categories, self-classified by the agencies that submitted them, mostly British, mostly well funded. The chart built on that split has no readable values and publishes no cell counts.

The peer-reviewed picture is thinner still. A systematic review of emotion in business-to-business contexts, searching two databases across fifteen years, retained 38 articles, of which the great majority use interviews and questionnaires, with attitudinal outcomes. The one empirical study closest to the question measures brand engagement tendencies and advocacy, not purchase, and reports an interesting inversion: the more a buyer says they rely on facts, the better they respond to emotional messaging, with the effect strongest among senior executives.

Evidence base for the claim that emotional advertising outperforms rational advertising in business to business marketsEvidence base for the claim that emotional advertising outperforms rational advertising in business to business markets. The commercial report most often cited for this claim was published in two thousand nineteen by a think tank funded by a professional networking platform, and its authors state directly that there are still relatively few business to business cases in the databank so sample sizes are small at less than fifty cases, that those cases may not be typical of business to business marketing in general because the databank is biased toward effective campaigns since all were entered into an effectiveness awards competition, that geographically they are skewed toward the United Kingdom although nearly a third come from elsewhere, and that they tend to have relatively big budgets. On classification, the same report states that case study authors, who in the case of business to business cases will be the business to business agencies and their clients, assess how important rational and emotional consideration are in purchase decisions. The chart built on that classification publishes no readable axis values and no cell counts, so with fewer than fifty cases split across categories each bar rests on a few dozen cases at most. The peer reviewed picture is thinner. A systematic literature review of emotion in business to business contexts searched two academic databases across the period from two thousand five to June two thousand twenty, screened an initial forty two papers plus eleven added by targeted search, excluded nineteen at second screening and added four from references, and retained thirty eight articles across fifteen years, of which eighteen percent are theoretical and eighty two percent empirical with qualitative and quantitative work in roughly equal parts and with interviews and questionnaires dominating the data collection methods. The closest empirical study to the question measures brand engagement tendencies and advocacy rather than purchase, and reports that the more a buyer says they rely on facts, the better they respond to emotional messaging, with the effect strongest among senior executives.The B2B evidence, in fullThe commercial reportunder 50 cases”biased towards effective campaigns""skewed towards the UK""tend to have relatively big budgets”Classified by the agencies themselves.The peer-reviewed literature38 articlesAcross fifteen years, two databasesInterviews and questionnaires dominateOutcomes are attitudes, not purchasesNone measures a contract won.
Fewer than fifty self-classified cases in the commercial report. Thirty-eight articles in fifteen years in the peer-reviewed literature, mostly attitudinal. Source : Binet and Field, The 5 Principles of Growth in B2B Marketing, The B2B Institute, 2019; Ribas and de Almeida, EMAC Proceedings, 2021 (2021)

What research with a published method actually found

There is a study that does what the database does not: it codes the advertising independently, against a published instrument, and links it to real sales.

The design: 2,317 television spots for 144 car models, over 45 months, in the United States. Content coded against a 118-item codebook, by twelve trained coders working in supervised sessions, with two coders per advertisement so that agreement could be estimated, and with allocation randomized. Informational and emotional content are measured as counts of cues, weighted by spend.

Two results matter more than the headline.

Informational and emotional content are not opposites. Their correlation across 2,317 spots is 0.01, with a probability of 0.67. The two are close to orthogonal, which means the whole framing of choosing one over the other misdescribes what an advertisement contains.

And which one works depends on the product. The effect of informational content on the sales elasticity was positive and significant for cars under 36,000 dollars and not significant above. The effect of emotional content was not significant below 16,000 dollars and positive above. By quality rating, information was significant below a rating of 3.7 and not above, while emotion was positive at any quality level.

The authors’ conclusion is conditional rather than universal: cheaper and lower-rated products should lean informational, premium products emotional. The study is observational rather than randomized, covers a single high-involvement category, and says so.

Conditions under which informational and emotional advertising content affected real salesConditions under which informational and emotional advertising content affected real sales, from a study published in the Journal of Marketing Research in twenty twenty one. The design covered two thousand three hundred and seventeen television spots for one hundred and forty four car models over forty five months in the United States, with the vehicles representing eighty two point six percent of category sales and the advertisements retained representing eleven point three billion dollars and ninety two percent of television investment. Advertising content was coded against a published one hundred and eighteen item codebook adapted from an established instrument, by twelve trained student coders working in supervised sessions of two eighty minute blocks with an author present, after a calibration phase on twenty five advertisements outside the sample, with each advertisement coded by two coders so that inter coder agreement could be estimated and with allocation and order randomized. Informational and emotional content are measured as counts of cues weighted by spend at month and model level. Two results matter more than the headline. First, informational and emotional content are not opposites, since their correlation across the two thousand three hundred and seventeen spots is zero point zero one with a probability of zero point six seven, meaning they are close to orthogonal and the framing of choosing one over the other misdescribes what an advertisement actually contains. Second, which one works depends on the product, since the effect of informational content on the sales advertising elasticity was positive and significant for cars priced under thirty six thousand dollars and not significant above that, while the effect of emotional content was not significant below sixteen thousand dollars and positive above it, and by quality rating information was significant below a rating of three point seven and not above while emotion was positive at any quality level. The authors conclude that managers of cheaper and lower rated products should favor informational cues while those of premium products should favor emotional cues, a conditional rather than universal recommendation. The study is observational rather than randomized, covers a single high involvement category, and states that caution must be exercised in extrapolating its results to other contexts.Not opposites, and conditionalCorrelation between informational and emotional content, across 2,317 spots0.01, with a probability of 0.67. The two dimensions are close to independent.Where each one had a significant effect on salesInformational contentsignificant under $36,000, not aboveEmotional contentnot significant under $16,000The method, so you can weigh it118-item codebook, twelve coders, two coders per advertisement, randomized allocation. Observational,single category, and the authors say so.
Coded against a 118-item instrument, two coders per spot, linked to real sales. The two dimensions correlate at 0.01. Source : Guitart and Stremersch, The Impact of Informational and Emotional Television Ad Content on Online Search and Sales, Journal of Marketing Research 58(2), 2021 (2021)

Two more things worth knowing before you commission either

Measuring the emotion does not tell you it sold anything. A study of 219 participants watching 64 award-winning commercials found that automatic facial coding explained 37 percent of the variance in self-reported joy, 25 percent of ad likeability, and 5.5 percent of the change in purchase intention. The authors add that they do not know whether that intention translates into purchase at all. Whatever such tools measure, it is closer to whether someone enjoyed the advertisement than to whether they bought.

And a rational claim carries a legal obligation that an emotional one does not. United States policy is explicit that advertisers must “have a reasonable basis for advertising claims before they are disseminated”, and that “as a matter of law, firms lacking a reasonable basis before an ad is disseminated violate Section 5 of the FTC Act”. Substantiation assembled afterwards does not cure the problem: “advertisers will not be allowed to create entirely new substantiation simply because their prior substantiation was inadequate.”

The obligation tightens sharply if you describe your evidence. Where a claim is express, using phrases like “tests prove”, “doctors recommend” or “studies show”, “the Commission expects the firm to have at least the advertised level of substantiation.” And implied claims count too, including ones made by imagery rather than words.

So the rational route is not the safe one by default. It is the one with a documentary requirement attached.

Substantiation obligations attaching to a factual advertising claim under United States federal policySubstantiation obligations attaching to a factual advertising claim under United States federal policy, as set out in the Federal Trade Commission’s policy statement on advertising substantiation of November nineteen eighty four and its later compliance guidance. The core requirement is that advertisers and advertising agencies must have a reasonable basis for advertising claims before those claims are disseminated. Objective claims for products or services represent explicitly or by implication that the advertiser has a reasonable basis supporting them, and a firm’s failure to possess and rely upon a reasonable basis for objective claims constitutes an unfair and deceptive act or practice in violation of section five of the Federal Trade Commission Act. The requirement of priority is a rule of law rather than a preference, since the policy states that as a matter of law firms lacking a reasonable basis before an advertisement is disseminated violate section five and are subject to prosecution, and that advertisers will not be allowed to create entirely new substantiation simply because their prior substantiation was inadequate. The level required depends on six factors, namely the type of claim, the product, the consequences of a false claim, the benefits of a truthful claim, the cost of developing substantiation for the claim, and the amount of substantiation experts in the field believe is reasonable. The obligation tightens where the advertisement describes its own evidence: when a substantiation claim is express, using phrases such as tests prove, doctors recommend or studies show, the Commission expects the firm to have at least the advertised level of substantiation, and where an advertisement implies more substantiation than it expressly claims the advertiser must possess the amount and type of substantiation the advertisement actually communicates. Implied claims are covered, including claims conveyed by imagery rather than by words, with published guidance giving the example of a brochure showing doctors in white coats, microscopes and a stack of medical journals as likely conveying an implied claim that a product has been clinically proven.The obligation attached to a factual claimBefore, not after”advertisers and ad agencies have a reasonable basis for advertising claims before they are disseminated”And you cannot fix it later”advertisers will not be allowed to create entirely new substantiation simply because their priorsubstantiation was inadequate.”Saying “tests prove” raises the bar to what you claimed”the Commission expects the firm to have at least the advertised level of substantiation.”
Before dissemination, not after. And saying tests prove raises the required level to whatever you claimed. Source : FTC Policy Statement Regarding Advertising Substantiation, 23 November 1984; FTC Health Products Compliance Guidance, December 2022 (2022)

What to do with all of this

Stop treating it as a binary. The one study that coded content independently found the two dimensions correlate at 0.01. An advertisement can be highly informational and highly emotional at once, and most good ones are.

Match the mix to the purchase, not to a slide. The evidence that exists is conditional: informational content did the work at lower prices and lower quality ratings, emotional content above them. In business purchases, where the buyer must justify the choice to colleagues, the informational half is not decoration, it is what your champion carries into the meeting. Both halves come out of the same place, which is deciding what the firm is entitled to claim it does better, before anyone writes an advertisement.

Do not cite the emotional advantage as a measured fact. It comes from a self-classified label paired with a self-graded outcome on the same form, the gap is unmarked for significance, and on the companion measure the order reverses.

And treat a factual claim as a document, not a headline. Hold the substantiation before publishing, at the level your wording implies, and remember that a picture of a laboratory makes a claim as surely as a sentence does.

The honest summary is that nobody has run the experiment everybody cites. No randomized trial has compared an emotional and a rational execution of the same product against real sales. The largest body of genuinely randomized advertising experiments, 389 split-cable tests on real purchases, concluded that its data “do not show a strong relationship between standard recall and persuasion copy test measures and sales effectiveness.” Which is a polite way of saying that the instruments used to declare a winner in this debate have not themselves been shown to predict one.