Segmentation of the Pharmaceutical Market in Ecological Marketing

Four AI models were used: ChatGPT-4o (W1), DeepSeek-V3 with Deep Think (R1) mode (W2), Qwen 2.5 Max Large Language Model (W3), and Perplexity PRO (W4). Traditional segmentation was performed using a complex faceted multifactor method. The obtained results were analyzed using the expert evaluation method based on the Likert scale, according to the following indicators: obligatoriness (A) and constructiveness (B) of the characteristic; necessity and sufficiency of the number of variables for the characteristic (C) and their constructiveness (D); segment attractiveness (E) and its alignment with the strengths of the implemented environmental measures (F).

When segmenting the market using the four AI models, 6 to 9 characteristics and their variables were selected (the total number of segments in each model ranged from 648 to 26,244). The highest frequency of positive expert evaluations for indicators A–D was observed in models W3 and W4. When selecting the key target segment (indicators E and F), segments with average values were prioritized, which, according to experts, was not always justified. Traditional market segmentation was conducted based on three characteristics (the most significant ones according to the AI models) and their variables, resulting in a total of 36 segments. The results of the expert evaluation demonstrated the comparability of data on similar key target segments between the traditional method and model W4.

The use of AI models for market segmentation allows for an expanded range of segmentation characteristics and criteria for selecting the key target segment. However, special attention should be paid to the accurate and precise formulation of the task (prompt) and the careful adjustment of the number of resulting target segments.