That is what happened to the trademark Wave Prazer. Filed in class 35, retail, there was no identical or similar registration in that class. Still, the application was refused: the only prior mark cited in the decision was WAVE, registered by DKT do Brasil in class 10, personal-care products.
Although the two marks were classified in different classes, INPI found that the products occupied markets close enough to create a likelihood of confusion, based on Art. 124, XIX of Brazil's Industrial Property Law (Lei da Propriedade Industrial, LPI, Law 9,279/96).
And the data shows how large this blind spot is. Across thirteen years of collision-based refusal decisions published by INPI, 9.08% of refused marks involved a conflict with a mark from a different class. Counting each collision pair, refused mark and cited mark, the share reaches 11.48%, roughly 1 in 9: 104,090 concrete pairs of refused mark and cited mark sharing no Nice class at all.
The Nice Classification organizes products, not markets
The Nice Classification splits goods and services into 45 classes for administrative purposes, and the most common prior-art search is limited to the class the applicant intends to file in. But Brazilian law does not confine collision to the same class. Art. 124, XIX of the LPI bars registration of a mark that reproduces or imitates one already registered for identical, similar, or related goods or services, where there is a likelihood of confusion or association. It is the word “related” that opens the door beyond the declared class.
INPI's Trademark Examination Manual points examination in that direction: the weaker the similarity between the names, the stronger the market relatedness between the goods or services needs to be to establish risk, and the reverse holds too. In practice, a trademark can lose to a competitor that would never show up in a search restricted to its own class, because that competitor's goods or services, however differently classified, occupy the same market space in the eyes of the consumer.
Chart 1
Cross-class conflict is not a rare exception
Share of collisions that cross Nice classes, measured two ways: by mark pair and by refused mark.
Source: TrademarkIQ, based on IPAS024 (refusal) decisions citing a specific application number, published by INPI between 2013 and 2026. Conservative criterion: any shared class already counts as non-cross-class.
The two numbers measure the same thing in different units. A mark can carry more than one collision citation in the same decision, which is why the per-pair rate (11.48%) runs a bit higher than the per-mark rate (9.08%). About 125,000 pairs were left out of the analysis for lack of a known Nice classification, almost all of them (99.83%) marks filed before 2000, under the system that predates INPI's full adoption of the international classification.
Where it happens most
The most frequent class pair does not involve retail: it is coffee, cereal, and confectionery marks (class 30) refused citing meat, fish, poultry, and dairy marks (class 29), 5,863 cases, consistent with how often these food categories share the same shelves and packaging.
From there, class 35, advertising, business management, and retail, shows up in most of the following pairs, against apparel, coffee and confectionery, technical and scientific services, cosmetics, devices and software, and on the other side, against financial services. That makes sense: retail and commerce, by definition, cross paths with nearly any product or service it sells. There is a second possible explanation for this concentration that this analysis does not isolate: class 35 also holds a disproportionate share of filings in Brazil, which on its own would make cross-references with it more likely, whether or not real market relatedness exists in each case.
Chart 2
The ten most-cited class pairs in cross-class refusals
Class of the refused mark → class of the cited mark, pair counts across the 2013-2026 decision history.
Source: TrademarkIQ, based on predict.v5_collision_history cross-referenced with the Nice classification, restricted to marks with a single class on each side of the pair.
Other real cases
Three other refusals, already final and with no pending appeal, illustrate the same mechanism from different angles. In all three, both the refused applicant and the cited-mark holder are companies.
The most telling one does not depend on similar names. METALFLEX PORTAS E JANELAS, filed in class 6, metal hardware and window frames, was refused citing three marks in the same decision: two with close names in the same class, FLEXMETAL and METALFLEX, and a third with no relation in name or product, SASAZAKI, registered in class 3, household cleaning products. The decision does not weigh each citation separately, but the logos help explain the third one: the same layout, a blue band with the name in white over a red band.
Visual similarity, not textual
Refused mark
Metalflex Portas e Janelas
Class 6 · Metal hardware and window frames · Metalflex Indústria e Comércio de Alumínio e Vidros Ltda
Application 912403543
Cited mark
Sasazaki
Class 3 · Household cleaning products · SSZK Empreendimentos Participações Ltda
Application 910233101
Different names, different products, and different classes. Even so, the mark SASAZAKI appears among the prior citations INPI relied on, and the visual similarity between the two marks is evident.
Refused mark
Telesul Internet
Class 38 · Telecommunications · Luiz Carlos Garcia & Garcia Ltda ME
Application 928159957
Cited mark
Telesul
Class 35 · Business management · Telesul Telecomunicações Ltda
Application 820131474
Both companies operate in the telecommunications sector. The collision crossed the formal classification, not the real market: one was registered in telecommunications services (class 38), the other in business management (class 35).
Refused mark
FazBem Veg
Class 40 · Material treatment · Esquelino & Araujo Ltda
Application 928160955
Cited mark
Faz Bem
Class 43 · Food service · Faz Bem Asa Norte Comércio de Alimentos Ltda ME
Application 911568824
Both operate in the food sector, at different links in the chain: one in product processing (class 40), the other in restaurant service (class 43).
What this means for anyone filing a trademark
Checking for collision only within the intended class leaves a meaningful share of the risk out of the analysis. That share sits in related classes, connected by product type, sales channel, or audience, even when the formal numbering is different.
That does not mean searching all 45 Nice classes for every name. It means mapping, before filing, which classes actually share a market with the intended activity, and giving them the same scrutiny given to the application's main class.
This is also one of the problems that motivated TrademarkIQ to build Predict: analyzing trademark risk beyond the formal boundaries of the Nice Classification, factoring in relationships between classes that a conventional search may not capture.
The class is a starting point for the search. It should not be its limit.
Learn about PredictHow we calculated this
Source and period. Refusal decisions (code IPAS024) published by INPI between July 30, 2013 and August 25, 2026, restricted to cases where the decision's supplementary text explicitly cites the colliding application's number (the “Processo NNNNNN” pattern). 606,700 marks and 1,031,632 mark/cited-mark pairs in that universe.
Cross-class criterion. A pair only counts as cross-class when the refused mark and the cited mark share no Nice class at all, across every class held by each. Any overlap, even partial, is classified as non-cross-class. This is a conservative criterion: it does not distinguish strong from weak relatedness, only the total absence of overlap.
Exclusions. 125,045 pairs (44,775 marks) were left out of the analysis for lack of a known Nice classification on at least one side. We checked the filing date of those marks: 99.83% predate 2000, consistent with older registrations that predate INPI's full adoption of the Nice Classification, not with a sampling gap.
Class pairs. Calculated only between marks with a single Nice class registered on each side of the pair, to avoid ambiguity in counting multi-class marks.
Named cases. Selected among refusals with a final administrative decision: more than 120 days since the IPAS024 publication with no subsequent decision (or, in Metalflex's case, a refusal status with no appeal filed, already on record with INPI itself), no pending appeal, and only when both parties involved are companies.
Logos. Images of the composite marks, drawn from the public INPI image repository that TrademarkIQ maintains, shown only for marks refused on a final basis and their respective cited mark, with no alteration.
Limitations. Class 35's concentration in the most frequent pairs is consistent both with real market relatedness and with the sheer volume of filings in that class; this analysis does not isolate the two explanations. A citation in the refusal decision does not necessarily mean it was the only possible ground for refusal, only what INPI recorded in the published text.
Sources
About the author
Ricardo Vinhas
Founder of TrademarkIQ
Ricardo Vinhas is the founder of TrademarkIQ and a specialist in strategy and data intelligence. He holds a bachelor’s degree in Statistics from the Federal University of Bahia and a master’s degree in Statistics and Information Management, with a specialization in Marketing Research and CRM, from the NOVA Information Management School, NOVA University Lisbon, where he also completed a postgraduate program in Digital Enterprise Management. In 2026, he completed the Certification in Intellectual Property Law & Strategy from the WIPO Academy and INPI.