From Many to Few: A Graph Theoretic Screening Tool for Product Developers
Ennis, J. M., Fayle, C.M. and Ennis, D. M. (2011). IFPress, 14(2) 3-4.
Michelangelo reportedly carved his many masterpieces by removing all that was irrelevant to his final goal. While this approach clearly benefited the artist, it can also serve the brand and product developer in search of best combinations of items. Whether these items are juices in a mixed-juice drink box, flavor combinations for savory snacks or topping choices on a pizza, the practical problem is often the same, namely that from a moderate number of items an astounding number of combinations can be formed. While a range of techniques, from group discussion to fractional factorials and conjoint analysis, are currently used to trim down the full list of combinations to a list small enough for targeted testing, no technique in common use is specifically built to address this problem. Currently, much depends on the category-specific expertise of the product developer with the risk that surprising but potentially viable combinations might be mistakenly excluded from consideration. In this report we address this problem by recommending a new approach, based on relatively young mathematical techniques, that recognizes the special structure of this problem and allows us to systematically screen down a large list of combinations to one of manageable size.
Figure 1. A graph showing pairwise appropriateness information for pizza toppings.