Design and Early Validation of an AI-Assisted Perfume Website to Reduce Purchase Uncertainty in Online Perfume Selection: A Prototype-Based Study of SCENTARA

Abstrak

Perfume is a sensory and experiential product whose purchase decision is often backed by direct sniffing, skin testing, and personal interpretation of aroma character. Consumers who buy perfume online typically face purchase uncertainty, since they cannot smell the products immediately and have difficulty interpreting fragrance language. This is particularly true for SCENTARA, a new Indonesian perfume brand inspired by the Nusantara identity. The goal of this study is to build and conduct an early validation of a prototype for an AI assisted perfume recommendation website for SCENTARA. The study applies a design oriented research strategy with a qualitative dominant mixed method design. Exploratory interviews were carried out with eight potential customers to uncover purchase uncertainty, fragrance language problems, and demands for website guidance. The prototype was then tested on 60 questionnaire responses, followed by a follow up physical sample expectation match validation with 10 selected respondents. The results demonstrate that the prototype was favourably accepted across all five assessment dimensions. Users liked the Scent Finder and the explainable recommendation logic because they made abstract scent preferences more comprehensible for product selection. However, the least favourable item level outcome was the item on lowering uncertainty without personally smelling the product, confirming that digital assistance may lessen but not completely remove uncertainty. The physical sample validation gave directional evidence that the recommendation flow provided a meaningful starting direction for selected high confidence respondents. This study concludes that SCENTARA should develop its website not as a typical product display format, but as a dynamic platform for discovery and conversion assistance. Managerial recommendations include strengthening recommendation transparency, improving scent references and scent performance information, adding social proof, and incorporating behavioural data to refine the AI assisted recommendation logic.