When consistency costs visibility: brand visual identity, category distinctiveness and algorithmic exposure — data and code

Brand management prescribes visual consistency: a recognisable and repeated visual signature is held to build identity and ease processing. On content platforms, however, an audience is not reached directly. A recommender system allocates exposure before any viewer evaluates anything, and what pays under human evaluation need not pay under algorithmic allocation. This study measures the visual consistency and the category distinctiveness of 126 size-matched video channels — 63 consumer-brand accounts and 63 independent creator accounts across four product categories — from 6,058 long-form video thumbnails, using multimodal image embeddings. Negative binomial models with channel-clustered standard errors show that visual consistency is negatively associated with views and that the penalty is concentrated among brand accounts, where a one standard deviation increase in consistency is associated with a 67% lower expected view count, while the corresponding association among creators is not distinguishable from zero. Category distinctiveness follows the predicted inverted-U among creators only. Decomposing the outcome shows the consistency penalty operates on exposure-weighted views and not on likes conditional on views, which locates the cost at the allocation stage rather than in viewer evaluation. Semantic-level and surface-level consistency measures could not be told apart empirically. The findings qualify optimal distinctiveness theory for algorithmically mediated settings and caution brand managers against transferring offline consistency prescriptions to platform feeds without adaptation.