Everyone in France remembers Maurice. Nobody knows whether he sold any chocolate.

01Central question and hypotheses

In 2001, Nestlé and the agency Ogilvy Paris released a thirty-second film directed by Claire Denis. In it, a four-year-old accuses his goldfish of having eaten all the chocolate. The line — « Tu pousses le bouchon un peu trop loin, Maurice ! », roughly « you're pushing it a bit too far, Maurice » — stuck. Twenty-five years later it is still cited spontaneously as the canonical example of funny French advertising.

That is precisely what makes it a useful object of study. Maurice is the case most often invoked to prove that advertising humour works. It is also a case where nobody knows what worked.

This study takes that tension as its starting point in order to address a question of direct interest to any brand with a constrained communication budget: does advertising humour still have conversion power in France today, or have we entered an era of more serious communication — and does this effect vary by region?

Four hypotheses tested

• H1 — Memory / efficacy dissociation. Memorisation of and attachment to old-school advertising humour remain strong through nostalgia, but do not predict the perceived efficacy of current advertising.
• H2 — Regional variation in humour type. The preferred type of humour (absurd, self-deprecating, wordplay, situational) varies significantly by region.
• H3 — Regional humour / seriousness gradient. A regional gradient exists in the weight given to humour relative to seriousness (utility, values, responsibility) as a lever of trust.
• H4 — Differentiated shift over time. That gradient has moved differently by region: some are said to have « kept » humour while others abandoned it faster.

One hypothesis out of four is confirmed, one is refuted, one is partially supported, and the last proves untestable with the chosen design. That final point is treated in section 12 as a finding in its own right rather than a measurement failure.

A prior reformulation

The initial framing of H3 and H4 assumed substitution: seriousness having taken the place of humour. The available literature requires that premise to be reformulated before it can be tested. The ADEME study conducted with Kantar Media/TNS Sofres and Adwiser, covering roughly eighty creative executions broadcast in 2008 and 2009 across television, outdoor, press and internet, concludes that humour can serve a serious message — by making it less guilt-inducing and more acceptable — rather than being mechanically displaced by it^[6].

The hypothesis under test is therefore not one of replacement but one of functional coexistence: do humour and seriousness occupy distinct roles in the decision chain, and do those roles vary geographically?

02Methodology and traceability

This study departs from Insaitia's standard pipeline. The object analysed is not a product market but an advertising phenomenon, which makes the market-sizing instrument irrelevant here. The design rests on two proprietary instruments and a layer of external confrontation.

InsaitFlash — comparative quantitative measurement

The protocol strictly separates two time frames in order to prevent any contamination between retrospective and current data. The two batches are never executed in the same run.

Because regional comparison is the analytical core of the study, panels were generated separately for each zone: a national run cannot be decomposed by region after the fact. The design therefore comprises ten runs — five zones × two batches — of twenty personas each, or two hundred persona observations.

InsaitTalk — qualitative exploration of mechanisms

The qualitative component targets the mechanisms behind the scores rather than rediscovering them: why Maurice remains memorable, whether yesterday's humour is transposable, whether regional difference is real or stereotyped, and what to tell an SME that is hesitating. Five open questions were put as a continuous conversation.

Scope was deliberately restricted to three zones out of five: France as a whole for the substantive synthesis, the PACA region because it carries the lowest historical external score, and Auvergne-Rhône-Alpes because no external data exists there — making it the zone where the proprietary qualitative contribution carries the highest informational value. Île-de-France and Hauts-de-France remain Flash-only for this wave.

A composition correction was applied to the Talk panel. The Flash Batch A had proved concentrated among 31–65 year-olds, with no voice under thirty — a problematic bias for judging the current efficacy of humour. The Talk criteria widen the age band, diversify purchase channels and require a broad spread of adoption profiles.

External confrontation layer

Panel results are systematically confronted with three independent families of source: the reference meta-analysis on advertising humour^[1], two general-population surveys carrying regional detail^{[3, 4]}, and the available French advertising rankings and post-tests^[5]. Question 3 of Batch B was worded as a direct echo of the BVA 2016 item (« in France, can one laugh at everything »), precisely to make that comparison possible.

03The object of study: three levels of success, one missing

Before measuring what the public retains of Maurice, it is necessary to establish what the sources permit one to assert about the campaign itself. The documentary research conducted upstream of the study yields a three-level distinction that then structures the entire interpretation.

What is confirmed

A collection record held by the Musée des Arts Décoratifs in Paris attributes the 2001 campaign to Nestlé, to the agency Ogilvy Paris and to the director Claire Denis. It states a duration of thirty seconds and describes the film as a genuine success^[2]. The same record mentions several awards: an Effie France, a professional distinction of the CDA type, and the « Coup de cœur du public » prize at the Pub et Humour de Paris festival. A secondary advertising archive additionally associates the film with an Effie France 2001 and a Gold award in the « Consumer Foods » category — information consistent with the museum record, but to be treated as secondary confirmation.

The character was reused two years later, in 2003, for another reference in the portfolio — which makes Maurice a case of a recurring mascot rather than an isolated spot.

What was not found

• Sales evolution for the product before and after the campaign.
• Any quantified revenue increase, or any market share movement.
• A before/after awareness score, or a memorisation score accompanied by its methodology.
• An advertising return on investment, or campaign cost set against incremental sales.
• Any measure of purchase intent or trial specifically attributable to the film.
• Any data permitting a quantitative comparison between the 2001 version and the 2003 revival.

Secondary sources sometimes assert that the advertisement had a major impact on consumption, but without producing data, a sample, a comparison period or an advertiser source. Such assertions cannot be treated as established fact. Likewise, attribution of a Cannes Lion or a major professional Grand Prix to the campaign could not be verified.

This absence is not a defect of the campaign: it is the study's point of departure. The case most frequently mobilised to demonstrate the effectiveness of French advertising humour contains, in the accessible public sources, no measure of commercial effectiveness. What remains is to determine whether that absence is an archival gap or the manifestation of a mechanism. What follows shows that it is the latter.

04Panel composition and acknowledged biases

The InsaitFlash panels rest on synthetic personas whose demographic and regional distribution is calibrated on a public reference set of 5,855 real French consumer profiles.

One limitation must be stated explicitly rather than minimised. That reference set is a food behaviour reference, not an advertising or media perception reference. The demographic and regional representativeness of the personas is therefore sound; their responses on the perception of advertising humour are, by contrast, model-generated extrapolations rather than human responses empirically verified on this specific subject. That is the rationale for the external confrontation layer described in section 02.

Each panel comprises twenty personas per batch, in strict parity of ten men to ten women, spread across five socio-professional categories.

Three structuring biases

Absence of the youngest respondents in Batch A. The national retrospective panel contains no voice under thirty, and the regional panels start between 19 and 22 depending on the zone. Yet that is precisely the generation that did not grow up with Maurice, and therefore the one that would allow attachment to advertising humour to be distinguished from attachment to one's own childhood. The Talk design was parameterised to correct this.

Single stimulus. Batch A is anchored on a single example. The regional reports themselves flag this as a confirmation bias risk, steering responses towards nostalgia rather than free exploration of advertising memory. Nostalgia scores must therefore be read as scores of agreement with a proposed stimulus, not as scores of spontaneous recall.

Batches probably disjoint. Age bands differ between Batch A and Batch B within the same zone — for example 21–65 against 18–64 in Hauts-de-France — indicating two distinct persona sets rather than a followed panel. The A → B comparison is therefore between-subject, not within-subject. This study cannot assert that the same individuals who remember Maurice also report buying on information; it can only establish that these two findings coexist at regional panel level. Every formulation that follows respects that constraint.

Finally, socio-professional composition varies appreciably across panels — the share of managerial respondents ranges between 5% and 20% by zone in Batch B. This becomes material in section 11, where the BVA survey documents a gradient in attitudes to humour tied to socio-professional category independently of territory.

05H1a — What Maurice left behind: a structured memory

Batch A measures what survives of 1990s–2000s advertising humour. The results converge across the five panels and are stably ranked.

Three lessons emerge from this table, and the third is the most important.

The memory architecture is dual and stable. The character comes first everywhere (75 to 95%), the jingle second (65 to 75%). A national panel respondent states the order spontaneously: first the character, then the jingle because it stuck in your head, and situation third. This hierarchy does not vary from one region to another.

The perception that yesterday's humour was more unifying is near-unanimous — reaching 100% in PACA — and respondents supply the explanation themselves, which is not comic but medial: everyone watched the same television and saw the same advertisements, and talked about them at school or at work. This point is picked up in section 13, where it supplies an alternative explanation for the « decline of humour ».

The last row of the table is the weakest of all, in all five panels. The effect of humour on trust or affinity towards the brand tops out between 50% and 70%, while every other Batch A finding sits between 65% and 100%. More revealing still: it is the only Batch A item whose methodological confidence level falls to MODERATE, and it falls there in three panels out of five — Hauts-de-France, Île-de-France and Auvergne-Rhône-Alpes. Everywhere else, confidence is rated HIGH.

In other words: the measurement becomes uncertain exactly where the effect becomes distant. This is not an artefact — it is the signature of a mechanism, for which section 07 provides the theoretical formulation.

One wording nuance should be flagged for reading the table. The PACA item is phrased positively — simple absurd humour reinforces trust in the advertiser — whereas the Île-de-France and Hauts-de-France items are phrased restrictively: humour humanises the brand but is not the principal purchase trigger. PACA's 70% and Hauts-de-France's 50% therefore do not measure exactly the same proposition, and the inter-regional gap on this row must be read with caution.

06H1b — What humour does not do: convert

Batch B measures the perception of contemporary advertising humour, with no nostalgic reference. The central result is clear and does not change direction.

The purchase trade-off is decided everywhere. Placed in a position of choosing between a humorous and an informative advertisement for the same product, the panels favour information in proportions ranging from 75% to 90%. No region inverts that hierarchy; only its intensity varies.

The verbatims are explicit about the nature of the trade-off. A Hauts-de-France respondent formulates the distinction precisely: humour catches the eye, but it is information that makes you buy — a funny advertisement stays in memory, but if it says nothing about the product it creates no trust. An Auvergne-Rhône-Alpes respondent gives the concrete example of a detergent: she wants to know whether it removes stains, not to laugh.

The distinction respondents make is not « humour versus seriousness ». It is « attention versus information ». Humour is not rejected; it is assigned to one function — capturing the gaze — and excluded from another — justifying the purchase. That is a role assignment, not a condemnation.

This point is decisive for the reformulation set out in section 01. Respondents do not describe seriousness substituting for humour: they describe a division of labour between the two registers. The qualitative component confirms this reading and specifies its condition, set out in section 14.

07The attenuation chain — verdict on H1

The results of the two batches reproduce a structure that the scientific literature describes independently, and that convergence constitutes the study's most robust finding.

What the reference meta-analysis establishes

Eisend's meta-analysis, published in 2009 in the Journal of the Academy of Marketing Science, aggregates 369 correlations from research on advertising humour^[1]. Its conclusions, verified at source for this study, are as follows.

• Humour significantly enhances attitude towards the advertisement, attention and positive affect.
• It also enhances attitude towards the brand and purchase intention.
• It significantly reduces perceived source credibility.
• Contrary to earlier reviews, there is no evidence of an effect on cognitions, positive or negative, nor on liking of the advertiser itself.
• The decline from lower-order to higher-order communication effects is particularly strong: the effect size of humour on attitude towards the advertisement is twice as large as that on attitude towards the brand.
• This impact has remained rather stable over the past decades.

That final point is picked up in section 13. The fifth is the one that structures the entire reading: the effect of humour does not disappear, it attenuates at every step that takes it further from the advertisement itself.

Superimposing the two datasets

The advertising effect chain comprises four distinct steps that it is essential not to conflate: memorisation of the scene or line, attribution to the brand, attitude towards the brand, then actual purchase. The panel results distribute across that chain as follows.

Two independent methods produce the same curve. An international meta-analysis aggregating 369 correlations and five French panels surveyed in 2026 describe the same decline: strong effect on immediate reaction to the advertisement, attenuated effect on the brand, null or negative effect on the declared purchase decision. The meta-analysis quantifies the attenuation between the first two steps as a ratio of one to two; the panels manifest it through a drop in scores and through a degradation of methodological confidence at the same location.

This parallel retrospectively illuminates section 03. The absence of public commercial data on Maurice is not merely an archival gap: it is consistent with what the science predicts. A campaign can reach an exceptional level of memorisation of scene and line without its effect on sales being demonstrable. The two findings do not contradict each other — they belong to two different levels of effect, and the second is structurally weaker than the first.

Verdict H1 — hypothesis confirmed. Attachment to yesterday's advertising humour is massive (75 to 95% depending on the panel) and does not predict the perceived efficacy of current advertising, where information prevails by 75 to 90%. The dissociation is observed in all five panels, it is consistent with the reference meta-analysis, and it is consistent with the absence of commercial data on the reference campaign itself. Recalling the constraint set out in section 04: the dissociation is established at panel level, not at individual level, the two batches resting on distinct persona sets.

08H2 — Does the type of humour vary by region?

Hypothesis H2 predicted significant regional variation in the preferred type of humour. The data refute it.

Self-deprecation comes first in all five panels. The range runs from 60% in Auvergne-Rhône-Alpes to 85% nationally, but the ranking never inverts. No region favours a register that the others reject.

The qualitative component is more categorical still, and it is so in all three zones where it was conducted. The national panel describes strict regional differences as a marketing cliché exaggerated by agencies, and explicitly recommends abandoning geographic targeting in favour of segmentation by consumption profile and behaviour. The PACA panel categorically rejects the concept of regional advertising humour. The Auvergne-Rhône-Alpes panel refutes rigid North/South regional determinism.

What actually varies

The three qualitative panels do not, however, conclude to total uniformity. They shift the variable, and that is the most interesting result in this section.

In PACA, rejection of the regional stereotype coexists with a very clear local rejection of humour perceived as mocking, arrogant or condescending — expressed by one participant as the observation that in the South people do not much like being laughed at, even good-naturedly. In Auvergne-Rhône-Alpes, the variable introduced is that of legitimacy: local humour is accepted only if it rests on genuine understanding of the territory rather than a stereotype applied from outside. One participant illustrates the nuance by noting that an advertisement playing on mountain-dweller self-deprecation lands better than a cheap joke.

Regional variation is therefore not a variation in comic taste, but a variation in sensitivity to the sender's posture. It is not the register that changes from one region to another: it is the threshold at which humour is perceived as coming from above rather than from alongside.

This distinction meets an ethnological study of the development, since the 1990s, of Breton self-deprecating humour circulated on postcards and clothing^[8]. The study describes it as a cultural production tied to commercial objectives and embedded in a local context — that is, as a constructed and commercialised identity rather than proof of a pre-existing psychological difference.

Verdict H2 — hypothesis not confirmed. The preferred type of humour does not vary significantly by region: self-deprecation dominates everywhere, secondary registers converge, and all three qualitative panels explicitly reject the notion of regional humour. What varies is tolerance of posture — condescension, sender legitimacy — not the nature of the comedy. On the basis of these data, speaking of a regional posture-sensitivity index is defensible; speaking of a regional comic temperament would not be.

09Red lines: the study's strongest convergence

One result deserves to be isolated, because it is at once the most consensual and the most directly operational.

Agreement sits between 80% and 95% in every panel, and the list of subjects is practically identical from one region to another. Religion appears in all five, racism and disability in four. Variation concerns the inclusion of politics or sexism, not the core.

One verbatim recurs in almost literal form across several zones — in France you can laugh at a great deal, but you must avoid racism, religion and overly sensitive subjects — which indicates a shared norm rather than a territorial sensitivity.

This result intersects directly with the question of social risk. Fear that social media amplify the impact of a failed joke reaches 85% to 100% depending on the zone, with complete unanimity in three panels out of five. A PACA participant gives the most direct formulation: today a failed joke can destroy a reputation in twenty-four hours.

Immediate practical consequence: humour compliance is managed at national, not regional, level. A brand does not need five tolerance charters; it needs one list, and that list is stable. This is a substantial simplification relative to the starting hypothesis, and it has direct budgetary value for an SME.

One measurement limitation should nonetheless be noted. These items involve self-report on socially sensitive subjects, and the Auvergne-Rhône-Alpes report itself flags a probable social desirability bias. The measured levels indicate a declared norm; they do not prejudge actual reactions to a given execution.

10H3 — The regional trade-off gradient

Hypothesis H3 predicted a regional gradient in the weight given to humour relative to seriousness as a lever of trust. The data show a real gradient, but of an amplitude and a nature that require the hypothesis to be strongly qualified.

The gradient exists, but it is narrow and unidirectional. Fifteen points separate the top from the bottom, on a scale where every value sits above 75%. No region places humour ahead of information; none comes close to the balance point. The variation concerns the intensity of a universally oriented trade-off.

Three precautions apply before attributing this gap to territory.

• Amplitude comparable to methodological noise. On panels of twenty personas, a 15-point gap corresponds to three respondents. It cannot support a strong inference, and the design provides no confidence interval permitting it to be tested.
• Variable socio-professional composition. The managerial share ranges between 5% and 20% across Batch B panels. The BVA survey documents a gradient in attitudes to humour tied to socio-professional category — but oriented in the opposite direction to what one would observe if composition explained the regional gap. The confounder therefore exists without resolving the question, which weakens the territorial interpretation without replacing it.
• Wordings not strictly identical. As noted in section 05, some items are phrased positively in one region and restrictively in another, which limits strict row-by-row comparability.

Verdict H3 — hypothesis partially supported. A regional gradient is measured, it is consistent with each zone's distinctive qualitative signals, and it is uniform in direction. But its amplitude (15 points), the panel size (20 personas) and the variability of socio-professional composition preclude treating it as an established territorial effect. The defensible formulation is one of intensity variation within a common national trade-off, not one of an autonomous territorial gradient.

11Confrontation with public surveys

Because question 3 of Batch B was worded as a direct echo of the BVA survey item, a quantified comparison is possible. This is the principal external robustness test for the synthetic panel.

The reference data

The BVA survey conducted for the regional press interviews 1,006 people representative of the French population aged 18 and over, recruited by telephone then surveyed online on 24 and 25 March 2016, with quotas and stratification by region and agglomeration category^[3]. It establishes that 97% of French people like to laugh, that more than eight in ten laugh every day, that 93% consider that humour brings people together and 88% that it allows serious subjects to be discussed. By contrast, only 53% consider that in France one can laugh at everything, against 46% who think the opposite.

The gap between Bretagne and PACA reaches 31 points — one of the few quantified regional anchors available on this subject in France. A YouGov survey of March 2019 covering 1,008 people additionally records that 76% of Île-de-France residents consider themselves funny, against 65% in the South-East^[4].

A strong convergence

The PACA case constitutes the study's clearest robustness signal. The region carries the lowest external score in France on declared tolerance of humour (38%). Independently, ten years later, the PACA InsaitFlash panel produces the highest bad-buzz fear score (100%), and the PACA InsaitTalk panel spontaneously surfaces an explicit rejection of mockery, including good-natured mockery.

Two distinct instruments, a decade apart, one on real population and the other on synthetic personas, designate the same region as the most cautious on humour — and the qualitative component supplies the mechanism the external quantitative survey could not capture: it is not laughter that is refused, it is the position of looking down.

Two divergences not to be smoothed over

Hauts-de-France. The former Nord-Pas-de-Calais-Picardie region carries the second-lowest external score (49%). Yet the corresponding InsaitFlash panel is, alongside PACA, the one expressing the fewest red lines (80%) and one of the lowest bad-buzz fear scores (85%). The external and internal hierarchies do not coincide in this zone.

Île-de-France. The Paris-region panel records the study's highest red-lines score (95%), whereas the YouGov survey finds there the highest proportion of people considering themselves funny (76%). The contradiction may only be apparent — considering oneself funny and tolerating a brand being funny are two distinct dispositions — but the available data do not settle it.

Limits of the comparison

• Misaligned administrative boundaries. The external surveys use the pre-2016 regional map. No direct comparison point exists for Auvergne-Rhône-Alpes, and correspondence with Hauts-de-France is imperfect.
• Distinct objects of measurement. The BVA item concerns the possibility of laughing at everything in general, not advertising humour. The YouGov item concerns a self-perception, not campaign reception.
• Competing gradients documented by the source itself. The BVA survey records that 58% of men consider one can laugh at everything, and that higher socio-professional categories are more permissive than employees and manual workers. Gender and social category gradients therefore coexist with the regional gradient, at comparable amplitude — which directly supports the qualitative recommendation to segment by profile rather than by geography.
• Sampling uncertainty. Regions differ in size and composition, and nothing establishes that these gaps reflect stable regional humour styles.

Outcome of the confrontation: one strong convergence, one clear divergence, one zone with no comparison point. This mixed result is the honest one. It partially validates the synthetic panel in the zone where the external contrast was sharpest, and it signals that the fine regional hierarchy produced by the panel should not be published as a ranking.

12H4 — A hypothesis this design cannot test

Hypothesis H4 posited that the regional gradient has shifted differently by region over time: some regions having kept humour, others having abandoned it faster. This hypothesis cannot be tested with the available data, and it is more useful to explain why than to produce an answer the design does not support.

Three obstacles, two of which do not depend on the design

The design is cross-sectional, not longitudinal. A single measurement wave was conducted. Batch A surveys the present-day memory of a past era; it does not measure that era. A memory collected today is a present-day reconstruction, subject to nostalgia bias — a bias that the single stimulus used in Batch A mechanically amplifies, as the regional reports themselves flag.

No French public time series permits the comparison. The documentary research conducted upstream concludes explicitly that no harmonised base exists allowing the share of humorous advertisements among campaigns broadcast or best ranked to be measured year by year and region by region. This absence is not specific to this study: it affects all French research on the subject.

Creative tone coding is not standardised. Public rankings do not code creative register according to a stable nomenclature: the same campaign may be classified as humorous, offbeat, emotional, complicit, absurd or feel good. Computing a « share of humour » would therefore require complete manual recoding and a prior operational definition of what constitutes a humorous advertisement.

What a valid test would require

The hypothesis is not invalid — it is premature. Two designs would allow it to be addressed, and deserve to be costed before being set aside.

• Longitudinal route. Repeat Batch B identically across the same zones at regular intervals, preserving the exact question wording. The first wave then produces no temporal conclusion, but it constitutes the reference point from which subsequent waves will produce one. This is the cheapest and most robust route.
• Coded corpus route. Assemble a historical corpus of French campaigns, manually recode it against a stable and explicit register nomenclature, and cross it with available regional broadcast data. This route would measure advertising supply — what brands did — where Batch B measures perception — what the public believes they did. The gap between the two is precisely the subject of the next section.

Verdict H4 — hypothesis not testable as it stands. A cross-sectional design cannot establish a temporal shift differentiated by region, and no French external data compensates for that limitation. Any assertion to that effect in this study would be extrapolation. Wave 1 does, however, constitute a usable reference point for a later test, provided the question wordings are frozen now.

13Have we entered an era of seriousness?

This is the second half of the central question, and it is where the gap between measured perception and available sources is widest.

What the public perceives

The perception of a decline is massive and stable: 85% to 95% of respondents consider that brands use less humour than fifteen or twenty years ago, across the four regional panels where the item was measured. The sentiment is sometimes accompanied by a spontaneous explanation — brands are said to be afraid of getting it wrong.

What the sources show

• No French public statistical series demonstrates this decline. Trade documentation describes an evolution in tone, but that evolution is not measured by any comparable long series.
• The professional narrative is one of alternation, not decline. Strong presence of humour in the major spots of the 1990s and 2000s; rise of responsibility and brand purpose in the 2010s; increased caution during the Covid period; then a return of lightness in 2023–2024, attributed by professionals to an anxious and inflationary context^[7].
• Recent rankings contradict the idea of an erasure. The 2023 winning campaigns are described as highly humorous, light or offbeat; the 2024 edition remains marked by lightness, nostalgia, complicity and cultural references — responsibility and inclusion themes being, for their part, absent from that ranking^[5].
• Humour and seriousness are not mutually exclusive. The ADEME study concludes that humour can help address serious subjects and encourage action, on condition that it does not induce guilt or coerce the audience^[6].
• The meta-analysis concludes to stability. The impact of advertising humour has remained rather stable across the decades studied^[1].

An explanation consistent with both datasets

The perception of decline is therefore near-unanimous, and the decline itself is demonstrated nowhere. Rather than arbitrating between the two, it is more productive to note that the panels themselves supply an explanation of the gap, in Batch A rather than Batch B.

Between 75% and 100% of respondents consider that yesterday's humour was more unifying — and the reason they advance is never comic quality. It is always structural: everyone watched the same television, saw the same advertisements, talked about them at school or at work; today everyone sees different content, humour is more fragmented, more niche.

A reading hypothesis: what declined may not be humour, but the shared audience. A consumer exposed to personalised feeds mechanically encounters less advertising humour that others have also seen. The disappearance of the collective experience can produce the sensation that the object has disappeared, without the volume of humour broadcast having necessarily fallen.

This hypothesis is consistent with all the sources — it reconciles a near-unanimous perception of decline, an absence of documented decline, and a return of lightness observed by professionals. It is not demonstrated by this study, and is presented as such. It is, however, testable: it would suffice to cross a coded campaign corpus with media exposure data by cohort, following the second route described in section 12.

The risk factor

One element escapes this reasoning and must be treated separately: fear of bad buzz, measured between 85% and 100% and unanimous in three panels out of five. Trade documentation describes the mechanisms — immediate commenting, removal from context, viral repurposing — and an academic publication recommends favouring affiliative humour, which brings people together, over aggressive humour, which can turn against the brand in an unsuited context^[9].

No statistical series, however, allows this risk to be quantified: neither the number of campaigns withdrawn after a controversy, nor the share of campaigns subject to controversy, nor the effect of those controversies on sales. The factor is therefore credible and qualitatively documented, but unmeasured. It plausibly explains the caution described by professionals; it does not prove a change in volume.

14What the qualitative component adds: the coherence condition

The three InsaitTalk conversations converge on a single mechanism that the quantitative component could not surface, and which constitutes this instrument's distinctive contribution.

Maurice is not remembered for being funny

The national panel shifts the interpretation of nostalgia towards an unexpected reading: Maurice is perceived there as a symbol of resistance to abusive marketing, a form of plain common sense set against advertising imposture. One participant gives the sharpest formulation — the character embodies resistance to absurdity, and that is what is timeless. The description is explicitly attached to the refusal of imposture and to product authenticity, not to comedy.

The PACA panel proposes a complementary mechanism: the character's memorability comes not from situational comedy but from the humanity supplied by his acknowledged flaws. One participant sums it up by observing that he had a real flaw, that he exaggerated, which made him funny but also recognisable — not a perfect character selling you something.

The Auvergne-Rhône-Alpes panel formulates a triptych: universality of the human behaviour represented, absence of caricature, and social proximity. One participant describes Maurice as someone you would meet at the market or in a restaurant, someone from your own world, not an artificial creation.

Three independent panels converge: what survived twenty-five years is not a joke, it is a character. The operational implication is direct — a humorous campaign built on a gag is structurally less durable than one built on a character with a recognisable flaw.

The condition that governs everything else

On the question of reusing a character — Maurice having in fact been brought back two years later for another product — the three panels set the same condition, and set it as a strict one.

This condition illuminates a meta-analytic result that might have seemed paradoxical. Humour significantly reduces perceived source credibility^[1]. The qualitative component suggests where that credibility cost is actually paid: not with every use of humour, but when a gap appears between the humorous register and the substance of the product. The condition set by the panels — that the product delivers on its promises — is precisely the condition that neutralises the cost measured by the meta-analysis.

15Answering the central question

The question posed had two halves. Both call for a qualified answer, but neither calls for an ambiguous one.

16Operational implications

These results translate into concrete decisions for a brand operating under a constrained communication budget — the audience for which this study is intended.

What to measure a humorous campaign on

The most direct consequence is a change of indicator. A humorous campaign should be assessed on the three dimensions that French post-test instruments actually use — emergence (is the campaign noticed), attribution (is it connected to the right brand) and appreciation — and not on a sales variation that none of the available instruments isolates reliably.

The attribution step deserves particular attention, and it is the blind spot of the Maurice case as of this study's design. The line survived twenty-five years; that the brand and the product survived in the same memory is established nowhere. A campaign can pass step 1 and fail step 2, and that is precisely the scenario in which the investment is entirely lost.

Format choices

• Self-deprecation is the only format with cross-regional consensus (60 to 85%). It is the rational default in the absence of prior testing.
• Build a character with a recognisable flaw rather than a gag. The three qualitative panels converge: what lasts is an imperfect, socially proximate character, not a funny situation.
• The memory architecture is dual: character first (75 to 95%), sonic signature second (65 to 75%). A design that neglects the second forgoes the spontaneous recall vector.
• A single national list of subjects to avoid: religion, racism, disability, and depending on the zone politics and sexism. Compliance does not regionalise.
• Favour affiliative over aggressive humour on social channels^[9], consistent with the rejection of condescension observed in PACA and of illegitimacy observed in AURA.

Territorialisation

The recommendation is counter-intuitive and convergent across the three qualitative panels: do not build regional humour targeting. The territorial stereotype is explicitly rejected as a marketing cliché. If a brand nonetheless insists on a local inflection, the condition set by the panels is one of legitimacy — local humour is accepted only if it emanates from fine-grained understanding of daily life, not from a stereotype applied from outside.

A gap to know before investing

One sector data point deserves flagging to brands considering humour as a differentiation lever. A publication on Generation Z reports that 81% of consumers appreciate brands using humour, but that only 28% consider they do it well^[5]. These figures come from a communication relayed on a professional network rather than from a complete methodological record; they should be verified at source before any public use.

Subject to that caveat, the gap is directly actionable: it indicates that humour is not a risky bet in itself, but a lever with a wide distance between expectation and execution. That is an argument for validating register and execution in advance, rather than relying on isolated creative intuition.

17Methodological limitations

These limitations do not affect the direction of the principal results: the memory/conversion dissociation is observed in all five panels, converges with an independent meta-analysis, and is consistent with the absence of commercial data on the reference campaign. They do, however, delimit what the study authorises one to decide — a choice of register, format and level of segmentation — from what it does not: a regional ranking, a temporal conclusion, or a forecast of effect on sales.

18Conclusion

The Maurice case is useful precisely because it disappoints. The French advertisement most frequently cited as proof that humour works is carefully documented as a creative and cultural object — author, agency, director, duration, awards — and contains, in the accessible public sources, no measure of commercial effect. No sales, no market share, no return on investment, no comparison between the 2001 version and the 2003 revival.

That absence might have been nothing more than an archival gap. The panel data show that it is consistent with a mechanism. Five French panels surveyed in 2026 reproduce exactly the curve that a meta-analysis of 369 correlations has described since 2009: strong effect on memorisation and attention, effect halved on attitude towards the brand, null or negative effect in the declared purchase trade-off. The most telling detail is methodological — the only item in the retrospective component whose confidence falls to MODERATE is the one measuring the most distant step. The measurement becomes uncertain exactly where theory predicts the effect weakens.

On the regional dimension, the study refutes more than it confirms, and that is a finding in itself. The preferred type of humour does not vary: self-deprecation dominates everywhere. Red lines do not vary either: the same short list appears in all five panels, at between 80% and 95% agreement. The only gradient measured is one of intensity within a universally oriented trade-off, and its amplitude — fifteen points on panels of twenty personas — does not support a ranking. The three qualitative panels go further and explicitly reject regional humour as a cliché sustained by marketing itself, recommending segmentation by consumption profile rather than by geography. Confrontation with the public surveys confirms that caution: strong convergence in one region, clear divergence in another, no comparison point at all in a third.

On the question of an era of seriousness, the answer is probably that it did not happen. The perception of a decline in humour is near-unanimous in the panels, and that decline is demonstrated by no French public series. The professional narrative describes alternation rather than decline, recent rankings remain dominated by lightness, humour proves compatible with serious messages, and the meta-analysis concludes that the effect has been stable across decades. The panels themselves supply the explanation of the gap, and it is not comic but medial: what declined may not be humour, but the shared audience. A joke seen by everyone at the same time produced a collective experience that personalised feeds no longer produce. The object may have remained; it is the sharing that disappeared.

There remains the condition that the qualitative component surfaces in all three zones where it was conducted, and which governs everything else. What allowed Maurice to cross twenty-five years is not his funniness but his character — an acknowledged flaw, an absence of caricature, a social proximity. And what entitles a brand to use humour is not its creative budget but the absence of any gap between what the character promises and what the product delivers. The meta-analysis establishes that humour significantly reduces perceived source credibility; the panels indicate where that cost is paid. It is not paid with every joke. It is paid when the joke promises more than the product.

For a brand, the translation is simple and amounts to an inversion. Humour is not what makes people buy: it is what makes people look. Measuring it as a conversion lever condemns it on a criterion it does not meet; measuring it on emergence, attribution and appreciation judges it on what it actually does — and what it does, it does with an effect size twice that of the following step.

19Sources and references

Scientific literature

• ^[1] Eisend, M. (2009). A meta-analysis of humor in advertising. Journal of the Academy of Marketing Science, 37(2), 191–203. DOI: 10.1007/s11747-008-0096-y. — 369 aggregated correlations. Central reference for this study, verified at source.
• ^[8] Blanchard, N. (2012). Tout est bon dans le Breton. Ethnologie française, 42(4), 697–706. — Regional humour as cultural and commercial construction.
• ^[9] emlyon business school (2025). Aggressive or benevolent: which form of humour to favour on social networks? — Recommendation of affiliative humour.
• Questions de communication (2006). Psycho-linguistic study of the effects of humour in advertising. — Reception depends on comprehension of the comic mechanism, on context and on audience characteristics.
• Psychologie et éducation. Secondary school students' sensitivity to humour in television advertising.

Institutional sources and public surveys

• ^[2] Musée des Arts Décoratifs, Paris (2017). Nestlé — Chocosui's — Maurice, 2001. Collection record. — Attribution, duration, awards.
• ^[3] BVA for the regional press (April 2016). Les Français et le rire. n = 1,006, representative of the French population aged 18 and over, telephone recruitment then online interviewing on 24 and 25 March 2016, quotas and regional stratification. Verified at source.
• ^[4] YouGov (March 2019). Survey on considering oneself funny. n = 1,008.
• ^[5] Ipsos France. Palmarès de la Pub du Grand Public (with FranceTV Publicité, since 1986); Le monde en chiffres — Les Français et la publicité; Les Gen Z et la pub. — The last two figures cited require verification against the original methodological record.
• ^[6] ADEME, Kantar Media/TNS Sofres and Adwiser (2011). Study on humour in sustainable development communication strategies, approximately 80 creative executions analysed (2008–2009).
• AdsSpot advertising archive. — Secondary confirmation of the awards.

Trade press

• ^[7] Stratégies (2024). L'humour de retour dans la pub française.
• Stratégies. Peut-on encore rire dans la pub ? — Increased caution during the Covid period.
• Stratégies. Le rire : une ficelle efficace à manier avec parcimonie. — Distinction between memorisation, awareness and purchase intent.
• Kantar Media. Local advertising on television and radio. — Distribution of regional advertising activity, not perception of humour.

References ^[1] and ^[3] were verified directly at source for this study. The others come from the documentary work conducted upstream and should be subject to a check of links, dates and identifiers before final publication. None of these sources is the origin of the panel data presented, which comes exclusively from the two instruments described in section 02.