2. ITEM RECOMMENDATION: The approach should name specific categories which might be suitable for customized promotional campaigns in each segment. The recommended categories
4.5 Profitability Simulation
4.5.1 General Assumptions
We can start to answer the questions posed above by considering the major objectives of promotional campaigns in general. In retailing, price is the most common variable used to distinguish a retailer from his competitors. Most marketing campaigns refer to the prices of featured categories or products (cf.
Blattberg and Neslin 1990). Retailers usually have two main reasons for lowering the prices of specific goods:
• First, they often assume price-promotional elasticities of demand with a value lower than −1 concerning the featured products. This means that price reductions incite customers to buy more of these products, with earnings compensating or exceeding the losses (cf. Blattberg and Neslin 1990, van Heerde, Leeflang and Wittink 2004).
• Second, lowering the prices of certain categories should boost the visiting frequency of customers.
If potential customers enter the store, they usually buy not only the promoted items but also ad-ditional products from the entire assortment. Managers expect from this strategy that the sales made by new or regular customers in the other categories will compensate for the loss involved in reducing the price. An example of this strategy is the selling of loss-leaders (cf. Walters and Rinne 1986, Blattberg and Neslin 1990, Fox and Hoch 2005).
Consider recent promotional campaigns conducted to achieve these goals by featuring specific products or categories. One of the most popular techniques is to highlight the “bestsellers” (top or hit products) of a company (cf. Brijs et al. 2000, Bodapati 2008, Elberse 2008). The question arises, how do retail managers select the items which are to be used for those campaigns? A wide variety of methods for defining the items exists, depending on e.g. the calculation of specific operating figures concerning the profitability of the categories. Nevertheless, we assume that most marketers choose the promoted categories simply according to personal experience. A common technique to mark the interesting categories is to use
selection queries in the transaction or sales data. Here, we consider three simple queries determining the Φitems heuristically. The query can select theΦitems in the entire sales data (i.e. HFC + LFC) which generate the highest values in the following areas:
• profit margins (option 1)
• revenue (option 2)
• purchase frequencies/support values (option 3)
Featuring the most profitable categories seems intuitively to be the appropriate method (option 1): the re-tailer wants to boost sales in items which contribute the largest share to his profit. However, according to our stated objectives of promotional campaigns, the frequency of customer visits will probably increase if the retailer features the categories occurring in the highest share of all transactions (option 3). Promot-ing the items with a high support value might attract more customers and could initiate cross-sellPromot-ing most effectively. Concerning the first option, the retailer has to consider that the most profitable items do not necessarily also show a high support value. In practice, the decision maker could also feature the items generating the highest revenue since this is a common target figure in retailing (option 2). Table 4.7 lists for each of the three options theΦ= 4 categories extracted from months one to ten of the second sample.
The highest profit is achieved with sausages and the highest revenue with bottled beer. Various vegeta-bles are found most often in single market baskets. All three options define categories which belong to the HFC.
Option 1 (profit margins) Option 2 (revenue) Option 3 (support) Cat. 1 Types of sausage Bottled beer Other vegetables Cat. 2 Del. sausage/meat products Del. sausage/meat products Whole milk Cat. 3 Other vegetables Fizzy drinks Delicacy bakery
Cat. 4 Delicacy bakery Other vegetables Del. sausage/meat products
Table 4.7: Result of the three options to selectΦ= 4 categories heuristically
Table 4.8 shows the number of transactions including at least one of the corresponding items of Table 4.7 as well as the profit which is achieved with these categories in months one to ten of the second sample.
The first value might be a useful indicator for the reach of the marketing campaign. The more purchase occasions are affected by the corresponding promotion, the more customers will probably be attracted to the store in the future. Notice that the calculation of the profit values does not include any assumptions about cross-selling or correlations with other items. As expected, the categories of option 1 achieve the
highest profit but show the lowest support values. The opposite is true for the categories of option 3.
Promoting the four categories with the highest revenue leads to values which lie in between.
Each option takes the stated objectives of a promotional campaign into account with a different empha-sis: promoting theΦ most profitable items from months one to ten will probably produce the highest increase in profit in the corresponding items. On the other hand, these items will not occur in the highest share of the transactions, as shown by Table 4.8. Concerning the stated objectives of typical price promo-tion, even retailers who implement customer-unspecific heuristics want to increase cross-selling and the frequency of visits. Featuring the items with the highest support values means that these items appear in more transactions than do other items. Hence, more households would be reached by the corresponding marketing campaign and the chance of cross-selling or unplanned purchases would rise. Nevertheless, our calculated profit achieved with these four items in the first ten months is very low (e 41,005.35). If the company uses theΦ items with the highest revenue, the categories generate a slightly lower profit compared to those in option 1. However, the number of transactions including one of the four items is higher.
Heuristic query Option 1 (profit margins) Option 2 (revenue) Option 3 (support values)
Number of transactions 47,356 48,511 51,028
Profit over all K in e 75,904.48 63,858.69 41,005.35
Table 4.8: Number of transactions containing at least one of theΦcategories, and profit achieved with the different categories of the heuristic promotion method
To show the differences between standard promotion heuristics and personalized target marketing, we compare the simulated profit gain produced by the customer-unspecific promotion method when theΦ categories of the HFC are determined using the three query options, and a segment-specific promotion campaign implementing long-tail categories (i.e. LFC) from our data-driven framework. If theΦitems are determined according to the three simple queries (options 1-3), all customers are presented with a price reduction in the same promoted categories. In other words, this method does not single out cus-tomers according to their individual characteristics. Here, we call such aggregated advertising campaigns
“customer-unspecific promotion heuristics”, e.g. sending leaflets informing all households about special prices in theΦmajor categories.
The customer-unspecific promotion heuristic will be compared to a campaign which uses theΦ cate-gories derived from our data mining approach for two successfully implemented advertising techniques designed by Dr`eze and Hoch (1998): category destination programs (CDP) and cross-merchandising (CM). Category destination programs award the participants a price reduction for all purchases made
within theirΦdetermined categories. For example, Dr`eze and Hoch (1998) implemented a “baby club”
whose members get a coupon for a price reduction of 10 percent if they spend at least $100 in the related categories. These programs correspond to the results of our approach perfectly since we show in detail how to identify the members of the program and which categories have to be promoted. Category desti-nation programs can be combined with cross-merchandising techniques easily. The cross-merchandising technique uses the customers’ affinity for specific goods to direct them through the store to other items.
For example, Dr`eze and Hoch (1998) suggest combining the best-selling categories with less frequently purchased items, e.g. by placing a sign near the bottled beer recommending certain snacks which are arranged on another shelf at the store. It can be shown that sales within the targeted categories and the rest of the assortment grew in the observed period. The authors of the survey attribute this to the increase in the distance covered by customers walking through the store to reach the promoted items, and the corresponding increase in their exposure to the assortment.