Light buyers, small baskets

A shopping basket is the term used to describe a set of items purchased together on a single occasion (Mild & Reutterer 2003), irrespective of how many items were purchased, or how they were collected in-store (e.g. shopping trolley, basket, or hand-carry). The number of items bought (known as ‘basket size’) can be used to identify light and heavy shoppers in-store on a particular day, in the same way that purchase frequency identifies light and heavy buyers.

When we examine the sales of light and heavy brand buyers, we observe that the heaviest 20% generate 40-60% of brand sales, with the remainder generated by the lightest 80% of buyers (Report 42 & 54; also see: Habel et al. 2003; Rungie et al. 2002; Schmittlein et al. 1993; Sharp 2010). However, there is little evidence about how this pattern extends to shopping baskets. Given that many manufacturers set goals to increase basket penetration with their branded goods, and retailers reportedly wish to increase average basket size, some basic knowledge about baskets is useful.

This report examines basket size patterns, addressing the following questions:

1. How many items are in the smallest 80% of baskets, and largest 20% of baskets?
2. How many sales (units and revenue) do small and large baskets generate?
3. How do these patterns differ across retail types?

Basket analysis and big data

This report draws from a 2016 panel of approximately 60 thousand US households that purchased 90 million products across nine million shopping baskets.

We examined basket patterns across ten retail types (Table 1), including food versus non-food retailers (e.g. grocery stores and hardware stores), specialty versus non-specialty retailers (e.g. pet stores and discount stores), and small versus large retailers (e.g. convenience stores and warehouse clubs). As retailers differ in many ways (average price, product offerings, location, and floor space, to name a few), studying this range of retailers is important to capture differences that may occur. The chosen retailers account for approximately 80% of trips (both small and large) in the dataset. The average number of unique stock keeping units (SKUs) per retailer was used to represent the overall size of the retail type.

Table 1: Retail types selected for analysis

Classifying small and large baskets across retail types

The many differences across retailers means that shoppers’ basket sizes will also differ – a small basket in a grocery store will likely contain a different number of items to a small basket in a convenience store. Therefore, we need to classify small and large baskets across retailers in a relative way. In-line with the Pareto principle, we classify large baskets (heavy shoppers) as the largest 20% in the dataset when ordered by size. Table 2 shows this large basket/heavy shopper classification, compared to average retailer basket size (as basket size distribution is right, positively, skewed, the median is the most appropriate average measure).

The results confirm that retail types differ by the number of items in a small and large basket. As average basket size increases across each retailer, so does the number of items that classify a large basket. For example, in grocery stores shoppers purchase nine items on average, the largest 20% of baskets have ≥23 items, and the smallest 80% have 1-22 items (0 items, i.e. browsing are not recorded in the data). In comparison, shoppers only purchase two items on average in hardware stores, and large ‘hardware’ baskets contain ≥5 items.

Table 2: Number of items in a large basket across retail types

How many sales do small and large baskets generate?

Now that the number of items in small and large baskets has been classified, it is possible to determine the proportion of sales they generate (Table 3). The results show that on average, the smallest 80% and largest 20% of shopping baskets each contribute approximately half of retailer unit sales. For example, in grocery stores, the smallest 80% of baskets (1-22 items) account for 52% of unit sales, and the largest 20% of baskets (≥23 items) account for the remaining 48%. This pattern occurs across retail types, despite the differing number of items in a small or large basket. So, even though large baskets in hardware stores only contain ≥5 items, these baskets still contribute 51% of unit sales.

For dollar sales, the largest 20% of baskets generate approximately 40% of revenue on average, and the smallest 80% generate the remaining 60% of revenue. Closer examination of each retail type shows that some retail types (e.g. grocery stores and discount stores) follow the 50:50 ratio, whereas other retail types generate more revenue from the smallest 80% of baskets (e.g. convenience stores and liquor stores). Retailers with a smaller average basket size typically generate more revenue from small baskets. For example, shoppers only purchase one item on average from convenience stores, and small baskets (1-3 items) account for almost 80% of revenue.

A notable deviation is observed in pet stores, which have fewer unit sales from small baskets (38%) that account for more revenue (71%). This suggests that pet stores are attracting few small, but expensive baskets (possibly due to high average item cost).

Table 3: Basket size Pareto share across retail types

So, how important are small shopping baskets?

This analysis tells us that both small and large baskets are important to retailers – on average they each account for half of retailer unit sales, and 60:40 percent of revenue respectively. Shopping baskets in retailers follow a ‘inverse J’ distribution (see Figure 1) – most baskets contain few items, few baskets contain many items, and the most frequent number of items purchased is typically one. Figures 2 to 4 at the end of this report show an example of the distribution in grocery stores, hardware stores, and liquor stores.

Figure 1: Aggregate basket size distribution across all retail types

Given that the bulk of shopping trips in any store will be smaller baskets, a goal of increasing basket size much seems rather heroic. It suggests that like brands, most growth will come from increasing the store’s number of shoppers. That to grow a retailer to grow should focus mostly on increasing ‘store traffic’, i.e. getting more shoppers (increasing penetration), being aware that many of these shoppers will only be buying a few items. Of course, this does not mean that retailers should neglect heavy buyers/large shopping baskets, as they still contribute 50% of unit sales, and 40% of revenue.

These findings highlight the importance of making it easy for shoppers to navigate and checkout in-store. If retailers are not efficient at servicing small basket buyers (who, of course may be large-basket buyers on another day), it affects a considerable proportion of shoppers, and may even have a detrimental effect on shopper penetration. Small basket shopping trips are short trips, shoppers quite reasonably are not expecting to have to spend much time in store.  Retailers need to fulfil this expectation, or lose shopping trips.  This is particularly relevant for retailers that have a smaller average basket size, as they rely more on small baskets for revenue. This also reinforces the attractiveness of categories that are likely to occur in small baskets, such as carbonated soft drinks, milk, confectionery, and alcoholic drinks (see: Tanusondjaja et al. 2016).

For manufacturers, the prevalence of short small shopping trips highlights the need to be easily recognisable, and physically available (in stock, and in a variety of stores and retail types). If a brand is not easy to find and recognise, it risks being missed by the considerable number of quick shoppers.

References

Dawes, J. G. 2016, ‘Brand Growth in Packaged Goods Markets: Ten Cases with Common Patterns’, Journal of Consumer Behaviour, vol. 15, no. 5, pp. 475-489.

Habel, C. A., Rungie, C. M., Lockshin, L. & Spawton, A. L. 2003, ‘The Pareto Effect (80:20 Rule) in Consumption of Liquor: A Preliminary Discussion’, Rungie, CM & Lockshin, L (eds), Australian Wine Marketing Colloquium, Adelaide.

Mild, A. & Reutterer, T. 2003, ‘An Improved Collaborative Filtering Approach for Predicting Cross-Category Purchases Based on Binary Market Basket Data’, Journal of Retailing and Consumer Services, vol. 10, no. 3, pp. 123-133.

Rungie, C., Laurent, G. & Habel, C. 2002, ‘A New Model of the Pareto Effect (80:20 Rule) at the Brand Level’, Shaw, RN, Adam, S & McDonald, H (eds), ANZMAC, Melbourne.

Schmittlein, D. C., Cooper, L. G. & Morrison, D. G. 1993, ‘Truth in Concentration in the Land of (80/20) Laws’, Marketing Science, vol. 12, no. 2, pp. 167-183.

Sharp, B. 2010, ‘Which Customers Matter Most?’, in Sharp, B (ed), How Brands Grow, Oxford University Press, Melbourne, Australia, pp. 39-55.

Tanusondjaja, A., Nenycz-Thiel, M. & Kennedy, R. 2016, ‘Understanding shopper transaction data: how to identify cross-category purchasing patterns using the duplication coefficient’, International Journal of Market Research, vol. 58, no. 3, pp. 1-12.

Key Ehrenberg-Bass Sponsor reports

Bogomolova, S., Anderson, K., Kennedy, R., Page, B., Sharp, A. & Sorensen, H. 2013. The Fundamentals of Shopper Behaviour. Report 64 for Corporate Sponsors. Adelaide: Ehrenberg-Bass Institute for Marketing Science, University of South Australia.

Romaniuk, J. & Wight, S. 2010. Do Your Heavy Buyers Stay Heavy, and What are They Worth? Report 54 for Corporate Sponsors. Adelaide: Ehrenberg-Bass Institute for Marketing Science, University of South Australia.

Sharp, B. & Romaniuk, J. 2007. There is a Pareto Law – but not as you know it. Report 42 for Corporate Sponsors. Adelaide: Ehrenberg-Bass Institute for Marketing Science, University of South Australia.

Analyses calculated (or derived) are based on data from The Nielsen Company (US), LLC and marketing databases provided by the Kilts Center for Marketing Data Center at The University of Chicago Booth School of Business. The conclusions drawn from the Nielsen data are those of the researchers and do not reflect the views of Nielsen. Nielsen is not responsible for, had no role in, and was not involved in analysing and preparing the results reported herein.

Additional Figures

Figure 2: Basket size distribution (Grocery stores)

Figure 3: Basket size distribution (Hardware stores) 

Figure 4: Basket size distribution (Liquor stores)