Unboxing Expectations through AI‑Driven Analysis of Fashion Reviews to Understand Consumer Satisfaction and Reduce Product Returns

The fashion industry faces deeply interlinked sustainability challenges that span environmental, economic, social, and technological dimensions. In fashion e‑commerce, short garment lifespans and high product‑return rates are particularly problematic. Returns signal misalignments between consumer expectations and experiences while generating substantial environmental and economic waste. This study investigates how experiential clothing attributes shape consumer (dis)satisfaction and return behaviour by analysing a large corpus of ASOS product reviews (n = 44.408). Using large language models, we analyse the presence, frequency, and sentiment of satisfaction attributes derived from Niinimäki’s Satisfaction Attributes for Clothing Longevity (SACL) and refine which attributes can be reliably detected in real consumer language at scale. The results show that consumer satisfaction is primarily driven by immediate, observable attributes, especially size, physical reactions, colour, and perceived value, while long‑term and use‑phase attributes remain largely absent from review data. Moreover, consumers describe product experiences in holistic and intertwined ways, complicating attempts to separate satisfaction attributes into clean analytical categories presented in theoretical frameworks. Based on these insights, the study proposes a three‑level set of design strategies that enhance review content (Level 1), translate this content, using AI, into structured insights (Level 2), and present these insights on product pages (Level 3) to reduce consumer uncertainty and return rates. Additionally, it also provides brands with actionable input for designing longer‑lasting garments. Overall, the findings demonstrate how AI‑driven review analysis can support more accurate consumer expectations and contribute to longer garment lifespans.