|
Abstract:
|
Food and Beverage R&D teams rely on multiple consumer metrics (e.g., mean overall liking (OL), %consumers with positive purchase intent (PI), and %consumers indicating that prototype meets their expectations (ME) relative to a product concept presented in the consumer test) to assess prototype readiness to move past development. These metrics are collected within the same questionnaire, and may become highly correlated due to halo effect. It is thus vital to obtain unbiased estimates of OL-PI, OL-ME, and PI-ME correlations (r) to justify the use of multiple metrics. To address this goal, a dataset with 13 products rated on OL, PI, and ME by 439 consumers via a crossover design was used. To remove the halo effect, the consumers were split into two: half was randomly assigned to Group 1 (G1) and the other half to Group 2 (G2). OL for each product was calculated from G1, while PI was calculated from G2, yielding r(OL,PI) at the product level. The said steps were repeated 5000 times to obtain a bootstrapped distribution of r(OL,PI). The above process was separately executed to obtain r(OL,ME) and r(PI,ME). Results will inform best practices on using multiple metrics in consumer tests.
|
