The effect of an incomplete block design on consumer segmentation
This presentation examines how balanced incomplete block designs – commonly used in wine and alcohol beverage studies to avoid intoxication, carryover, and fatigue – affect the ability to segment consumers into preference clusters. The researchers compared clustering outcomes from complete versus incomplete block designs using two approaches: clustering based on block effects (average responses) and clustering based on treatment effects (response vectors). Simulations showed that incomplete designs increase variance when clustering block effects, and when clustering treatment effects, consumer segments can only be reliably detected when clusters differ on more variables than the number of missing treatments per block. Despite these limitations, the findings confirm that incomplete block designs remain effective for data collection when practical constraints require them.
Browne, R., McNicholas, P., Castura, J. C., & Findlay, C. J. (2010). The effect of an incomplete block design on consumer segmentation. In: 9th Sensometric Meeting. 25-28 July. Rotterdam, The Netherlands.