Introduction: Descriptive analysis (DA) has been the most traditional tool for acquiring reliable sensory data. However, the need for rapid and low-cost techniques has enabled the consumer-based methodologies, such as check-all-that-apply (CATA). Considering inconsistencies of untrained assessors during evaluations of complex products, such as wines, training, even for a short duration, has been considered essential to improve data quality and reduce panel size. Objective: This research compared the evaluation of tropical red wines by panellists with different levels of training, uncovering how presentation of physical references, training period and reduction of panel size may affect the assessments. Methodology: Nine samples of Brazilian tropical red wines (Vitis vinifera L.) were characterized by DA with trained assessors (n=8), CATA with consumers (n=100) and CATA with semi-trained panellists (n=30). Terminology was instituted using the repertory grid method, and then had the vocabulary simplified, aiming to be used with consumers and semi-trained painellists. For DA, the 30h training was conducted in accordance to ISO 8586:2012. For semi-training, a two-hour period was performed in three steps: short class about tropical wines, definition of the sensory attributes and evaluation of the physical references. All samples were presented in a monadic way and balanced order. The similarity of sample configurations among methodologies was assessed using Multiple Factor Analysis and RV coefficient was calculated to quantify this similarity. Results: DA and CATA with semi-trained panellists provided the highest capacity of discrimination and the highest index of similarity for aged samples (95%). An improvement of 30% in the descriptive power of the samples, as well as an increase on attribute citation frequency was achieved by implementing a quick sensory training to consumers, also allowing for a reduction of 70% in the size of the panel; however, some misunderstandings related to technical terms were observed both in consumer and semi-trained variations. Conclusion: The two-hour training period had a positive effect on reducing the panel size and discrimination/characterization of samples, presenting itself as a valuable tool when time- and cost-efficient sensory profiling is needed. However, when technical results are required, consumer-based methods should not be regarded as replacement for DA.