Testing the robustness of brand partitions identified from purchase duplication analysis

ABSTRACT

Purchase duplication analysis examines the extent to which buyers of any brand A also buy other brands B, C, D and so on. A generalised finding from its use is that brands share their buyers with other brands approximately in-line with the size of those other brands. The approach is widely used by analysts and managers. One important use of the method is to identify partitions – brands that share buyers at a higher than expected rate. Partitions may form among competitor brands, but also among ‘same name’ or sub-brands that share a parent name (e.g. Coke, Diet Coke). A partition among same-name brands means they are cannibalising each other. Whether one’s focus is on cannibalisation within a portfolio, identifying close competitors, or to generally understand market structure, duplication analysis can provide insights. However, there are two potential confounds to its use: family buying and buying multiple brands on the same occasion. This study tests if these two factors confound the use of purchase duplications, using data from 12 grocery categories. The principal finding is that the identification of partitions is robust to these confounds. The study finds partitions among same-name brands are common and are also not due to these confounds.

KEYWORDS:

• Brand-switching
• purchase duplication
• partitions
• cannibalisation
• brand portfolio
• market structure
• sub-brands

Acknowledgements

I thank SymphonyIRI and Kantar for making the data available. All estimates or analyses in this paper based on SymphonyIRI Group, Inc. and Kantar data are by the author and not SymphonyIRI Group, Inc. or Kantar.

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

1. The partition for the Miller beer sub-brands was significant only at the p = 0.10 level in any year but was present in the initial analysis and for the following 3 years.

2. There are three Ragu brands, but the partition is more apparent between Ragu and Ragu Chunky. When we calculate the duplication coefficient for the three Ragu brands (Ragu, Ragu Chunky, Ragu Old World), the result is 1.7; among one-person households it is 1.4. Both these figures are higher than for the entire market (1.2 all households and 1.1 among one-person households). Therefore, the same-name partition holds when all three Ragu brands are used, and it holds among one-person households.

Additional information

Notes on contributors

John G. Dawes

John Dawes is an Associate Professor and Senior Researcher at the Ehrenberg-Bass Institute for Marketing Science. His research interests are in competitive market structure, brand performance metrics and the effects of price promotions on buyer behaviour.