Teppei Syokudo - Improving Store Performance: PPA Data Analysis Methodology
Data Exploration
Looking at both MW and RP’s product sales, the sale of Main - Meal has a decreasing trend, together with Main - Drink. This is likely due to the introduction of Set Menus, where customers tend to prefer purchasing sets rather than ala carte. The indirect relationship between Main - Meal and Set Menu is a lot more obvious in RP. We can see that the moment Set Menu was introduced in November, Main - Meal sales started dropping.
The most popular Main - Meal in both outlets would be the Kaisendon, which is sold 60 times and 50 to 60 times daily on average in MW and RP respectively. It’s relating set is the Seafood Feast which averages 40 and 24 times daily in MW and RP respectively.
We can also see that the sale of Main - Onigiri and Main - Fried have relatively stagnant to decreasing trends in both outlets. This means that the onigiris and fried items may not be very popular items. In order to boost sales of onigiris and fried items, Teppei Syokudo may want to consider introducing onigiri sets and fried item sets.
Literature Review
Market Basket Analysis was first introduced by Agrawal, Imielinski, and Swami in 1993. It aimed to identify when a customer purchases a particular item, a second particular item will be predictably purchased as well. Tan, Steinbach, & Kumar (2006) explains the methodology as follows. Given two items X and Y, a relationship in the form of association rules can be represented as {X → Y}. This suggests that when X is purchased, Y is also likely to be purchased. Support and confidence measures are used as threshold levels in association rules. With reference to the rule {X → Y}, support measures the probability of a transaction containing both X and Y while confidence measures the conditional probability of Y occurring when X occurs.
Historically, a classic example of Market Basket Analysis is the purchase of “beers” and “diapers”, two items seemingly unrelated but shown to have high association as they are often bought together. In recent times, common application of Market Basket Analysis can be found in online bookstores like Amazon, where customers are recommended “you may also like” books when they place a particular item in their shopping cart. In various academic literatures, Market Basket Analysis has been used to analyse purchase patterns in a multiple store environment (Chen, Tang, Shen, & Hui, 2005), identify ideal menu items (Ting, Pan, & Chou, 2010), and decide on appropriate product placements in a store (Charlet and Kumar, 2012). Applying Market Basket Analysis in an F&B settings, a restaurant may discover that customers tend to purchase food item X together with food item Y and drink Z. This information help managers to design product bundling strategies and helps floor staff cross-sell and upsell items successfully.
The set of items which meets the minimum support threshold are also referred to as Frequent Itemsets. There are various methods for generating Frequent Itemsets, the common ones being Apriori and FP-Growth, we will be using the latter in this paper. It is also worth noting that Zaki (2000) also introduced six algorithms for association mining - Eclat (Equivalence CLAss Transformation), MaxEclat, Clique, MaxClique, TopDown, and AprClique.
Figure 1. A Frequent Itemset Lattice
Consider the above lattice – each of these are itemsets. Algorithms have to identify the most efficient way to traverse the lattice and identify if a particular itemset is frequent. There are various ways of generating candidates for frequent itemsets and pruning, and this is determined by the algorithm used to carry out association analysis. The way the itemsets are generated and association rules created determine how computationally complex the analysis will be.
Therefore, considerations affecting the computational complexity of an algorithm have to be determined when dealing with mining association rules for large datasets. These include factors such as transaction width, number of products, minimum support level and max itemset size (Tan, Michael, & Kumar, 2005). Since the transaction width and number of products are predetermined, the team has chosen to specifically focus on the latter 2 factors to refine for our analysis - association thresholds and the max itemset size.
An important aspect of association analysis is the generation of frequent itemsets (or the elimination of infrequent itemsets). The minimum support (minsup) and minimum confidence (minconf) is taken into account. These are thresholds used to discover whether the itemsets in A --> B are frequent itemsets and whether A --> B is an acceptable association rule. While the team has explored algorithms to determine the optimal minimal support and minimal confidence levels such as applying Particle Swarm Optimization, the team has examined the data spread to determine appropriate minimum support and confidence levels.
Market Basket Analysis can be used to learn about customer purchase patterns so that customer facing staff can upsell and cross-sell in order to increase sales. Product bundles can also be created to increase attractiveness of products. Ideally, Market Basket Analysis seeks to identify interesting relationships between products in market basket transactions. This means that we seek association rules between products that have not been pre-determinately placed in purchasing relationships, such as items being part of a set or bundle. This is because naturally items within a set has already some form of association between them – an increased likelihood that these products are to be bought together. Therefore it is imperative that in our analysis, we handle items within a set and a la carte items differently. For items within set meals that are currently being offered, we use a subset of Market Basket Analysis to look at the effectiveness of the existing set meals and suggest areas for improvement to increase sales and demand. For a la carte items, we use Market Basket Analysis to identify suitable product combinations and cross-sell or upsell opportunities.
While there has been articles suggesting that association analysis methodology should incorporate item weights and transaction weights to better present analysis findings. In Weighted association rules: Model and algorithm (Ramkumar, Rankar, & Truc, 1998), the following example was suggested: “Caviar is an expensive and hence a low-support item in any supermarket basket. Vodka, on the other hand, is a high to medium-support item. The association: caviar => vodka is of very high confidence but will never be derived by the existing methods since the itemset {vodka, caviar} is of low support and will not be included.” While applying these improved algorithms provides benefits in terms of clarity of results, these advancement in methodology is excluded from the analysis as the variation in support levels are much less.
The “profitable-product death spiral” depicts that organizations or companies constantly attempt to ascertain their products’ profitability and because of pressures to cut costs and increase overall profitability, managers within the companies seek to remove the less profitable products. Cannon, Cannon, & Schwaiger (2012) claims that managers ignore “the fact that customers typically want an assortment of products, and that the deletion may weaken assortments that customers want. The resulting loss of sales weakens demand for previously profitable products, subsequently causing them to be dropped. This weakens the assortments even further, and so forth in a downward death spiral. By focusing on customer rather than on product profitability, marketers look at the portfolios of products their customers want rather than disrupting portfolios for the sake of individual product profitability.” Therefore, there is definitely worth in analysing the association between products amidst the need to ascertain product profitability.
