Selecting segmentation strategies for sensory acceptance: Evidence from multiple full-sample and reference-based methods using absolute liking and relative satisfaction

As the global fresh-cut market grows, meeting diverse preferences is critical to product design and positioning. In practice, sensory acceptance can be measured as absolute liking or as reference-anchored relative satisfaction (e.g., DOSD); the two provide complementary views. This methods-focused study structured segmentation choices by measurement and data transformation, and compared three pathways on a single dataset: (i) segmentation by clustering on absolute liking (9-point; raw and centered 9-point) using AHC (with CLV and K + 1), (ii) reference-based segmentation from DOSD relative satisfaction ( C summarizing the reference-relative satisfaction difference plus any reference-side bias, d ′ reflecting bias-free satisfaction discrimination), and (iii) a hybrid that clusters reference-anchored DOSD profiles. Acceptance data ( N = 110) were collected for eight lettuce cultivars (Korean and European types). In parallel, affect magnitude profiling (AMP) implemented with the double-faced applicability (DFA) test produced d ′ A sensory profiles; these were linked to acceptance via PLSR, and segment-specific drivers are presented as interpretable d ′ A profiles—consumer-term, SDT-based measures that quantify the magnitude of attribute applicability on a semantic-differential (bipolar) scale. Across pathways, absolute-liking clustering highlighted the overall liking structure; DOSD-based segmentation grouped consumers by satisfaction direction relative to the reference and was less sensitive to scale use; the hybrid combined anchoring with clustering to reveal profile-based subgroups. By explicitly detailing the response-bias implications of each pathway (scale use for liking; criterion/anchoring for DOSD), the study provided a practical decision framework for selecting full-sample (absolute), reference-based (DOSD), or hybrid approaches based on study goals and constraints, thereby supporting segmentation method selection and interpretation. • Compare liking clusters (AHC/CLV/K + 1), DOSD ( C , d ′), and a DOSD-based hybrid. • Mean-centered 9-point ratings reduces scale-use bias; clarifies sample effects. • DOSD segments consumers by satisfaction vs . reference; less scale-use bias. • AMP with DFA yields d ′ A profiles; acceptance drivers shown as interpretable d ′ A . • Fit-for-purpose framework for selecting full-sample, reference-based, or hybrid.