Teppei Syokudo - Improving Store Performance: PPA Findings
| Home | Product Portfolio Analysis | Improving Store Performance | Project Management | Documentation | The Team |
The analysis results are three-fold based on the breakdown of the analysis methodology: (1) analysis results of products that are already in a set in order to identify the most popular and unpopular components of the sets; (2) analysis results of products that are not within a set; (3) analysis results of product profitability.
Analysis Results of Products in a Set
The team found that certain drinks and side dishes are more likely to be bought together with a certain main dish than others. Recommendations based on the analysis results can be made as to whether the products should be retained in the product offering. Consider the following:
Recommendations can be made on which drinks are more popular with Main Dish X in a set – that “Ayataka” can be possibly removed from the set as well as by recommending highly associated drinks such as “Hot Green Tea” or “Cold Green Tea” will increase the satisfaction that customers derive. This is purely made on the assumption that the notion that “safety in numbers” is true. Stan (2016) suggests this is a belief based on the fact that a large number of consumers can’t all be wrong about the quality of value of the product combinations but “it’s entirely possible for a large number of people to be wrong, especially if few consumers research their options before making a purchase. As a result, a popularity appeal by itself might fail to convince savvy consumers.” It is then suggested that in order to make a quality recommendation is to back the popularity appeal with facts – in this case, it is that there is greater pleasure between “Main Dish X” and “Hot Green Tea” for example.
Analysis Results of Products Without a Set
With an acceptable minimum support threshold of 0.005 (since the average support for products is 0.02), the team discovered new association rules between products that not only provide a high association between the two products in the itemset, but also holds a substantial amount of support in the transaction data set.
The following recommendations can be made: (1) Providing discounts for either of Fried Dish A and Fried Dish B as well as Onigiri A and Onigiri B will see an increase in sales of the other. An alternative will be to provide sets that allows customers to choose these two as an option. Thirdly, the stores should consider continuing placing the products close together as this will increase the association between the products further. Ultimately, this recommendations will increase sales volume of the selected products. (2) A new set meal containing Main Dish A and Drink A can be recommended since there is an association between both products. Since it is already a popular choice to purchase the drink with this meal, placing them together in a set at a reduced price will increase sales volume of these two products further.
Analysis Results of Product Profitability
The profitability of products are ascertained using Cost of Goods Sold (COGS) and the pricing of the products. Under the assumption that products with lower profitability should be removed, the team has identified various products that have low profitability. One example is Onigiri B with a profitability of less than 26% (average for Onigiris are about 40%). In examining Onigiris to be removed from the product offering to save cost for the stores, managers should intuitively remove Onigiri B. However, with association analysis as we observed above, that there is a high association between Onigiri A and Onigiri B and the team uncovered that Onigiri A is in fact extremely profitable (profitability of 46%). By removing Onigiri B, a decrease in sales volume will be experienced by Onigiri A and hence a lowered overall store performance will be experienced.
Conclusion and Future Work
We have presented a new model of applying Market Basket Analysis in aiding a store’s product portfolio management by looking associations between products. This allows a store manager to have more targeted bundling options. Furthermore, the removal of the products based on the products’ profitability can now be analysed with consideration to the implications it has on other products.
Further research can be carried out on the actionable recommendations of Market Basket Analysis to F&B stores; examples could be menu management or the analysis of products being placed in a set. For stores with greater fluctuations in product support levels, an examination of a weighted model of Market Basket Analysis can be applied. However, in this regard it is imperative that the Delphi method is to be implemented to ascertain the weight of the product association; in other words, expert opinions from the sales staff or managers have to be harnessed or observed. Another likely advancement of the product portfolio management is to represent each product in a “social network”, where the size of the node is a factor of the product’s profitability and its importance based on the strength of association with other nodes and their respective importance. As such, a more accurate understanding of a product’s importance can by studied to prevent the “profitable-product death spiral”.
