The Economics of Cross-Selling

As consumers, we experience when retailers cross-sell routinely. It can be as simple as “Do you want fries with that shake?” or in more complex e-tail environments “Customers who bought X also bought Y.” Companies who do this well include Amazon and Netflix. In fact, Netflix thought the problem was so important that they setup a contest with a $1 million reward for anyone in industry or academia who can create and demonstrate a better recommendation engine for movies.

For distributors, growing existing customers is the overwhelming lifecycle management priority relative to customer acquisition and winning back lost customers as identified in an April 2011 survey conducted by Modern Distribution Management with 175 respondents (See Figure 1.) Up-selling and cross-selling are the two primary means of growing an account. Among these, cross-selling is paramount. Here’s why: many distributors have regular customers who buy the same set of 10 to 20 SKUs over and over, yet, they have thousands of other products that could be sold to those same customers.

However, customers segment the distributors from which they buy much in the same way that distributors segment customers, according to Jim Tenzillo, former VP Marketing at Grainger. For example, they know, that the supply company A is where they buy abrasives. They don’t think of supply company A for anything besides abrasives.

Distributors that are routinely successful at expanding the product categories sold to customers recognize that once a customer has been acquired, there is nominal incremental cost to sell other products. They go beyond selling the associated products recommended by the supplier and actually cross-sell other suppliers’ products. In addition to increasing the revenue per customer with each additional product category sold, they also increase the cost to the customer to switch to another distributor.

Barriers to Cross-Selling

The best salespeople generally know what to cross-sell based on their own experiences. While they are much better than average salespeople at cross selling, they are far from perfect. And since even small distributors have several thousand SKUs determining the best cross-selling relationships requires a more analytical approach than “eyeballing” the stock on hand.

The necessity of an analytic approach to discover cross-selling potential is best illustrated in a well-known and colorful cross-selling relationship now commonly used in retail settings. Using point-of-sale data, a large retailer found that diapers and beer are sold together frequently. In fact, this relationship was so strong that they decided to place diapers near beer in the store to maximize the cross-selling relationship.

One common explanation for this cross-selling rule is that a husband has been sent to buy diapers and decides to also pick up beer while at the market. In the database marketing seminars that we give, when we ask the audience to explain this phenomenon, we hear other much more piquant explanations. After the revelation and explanation about diapers and beer, it makes sense. Yet because it’s not intuitive, no retailer figured this out prior to the use of analytic techniques.

Identifying Good Cross-Selling Rules

Given the large number of products at most distributors, there are potentially many cross-selling rules that can be identified. The trick is to identify rules that when applied will produce a meaningful sales lift. We look at two attributes of good cross-selling rules:

• Support – This defines the percentage of all transactions in which items A and B appear together. If items A and B appear together frequently, then the combination may be a good candidate for cross-selling.
• Confidence – This defines the percentage of transactions in which item B appears where item A also appears. If item B appears in a high percentage of the transactions where item A appears, then the combination may be a good candidate for cross-selling.

For items A and B to be a good candidate for cross selling, there should be both high support and high confidence. That means that items A and B occur together frequently and that if item A has already been selected for purchase, there is a good chance of selling item B. For distributors with many SKU’s, high support often means that items A and B may appear in the same transaction in as few as half a percent of all transactions. The confidence from A to B should be much higher, typically more than 20 percent, because otherwise there will be an attempt to cross-sell item B when there is a low probability of being sold with item A.

A simple example for a restaurant will illustrate the concepts of support and confidence for guacamole and chips.

• 10 percent support for guacamole and chips
• 50 percent confidence from guacamole to chips
• 25 percent confidence from chips to guacamole

This means that 10 percent of all transactions have guacamole and chips. In addition, 50 percent of the time when customers buy guacamole, they also buy chips. However, only 25 percent of the time when they buy chips do they buy guacamole because they might also buy salsa or other condiments. The requirement for support can be reduced significantly as long as the confidence in the rules is high.

Identifying the products for cross selling is only the first step in developing cross-selling rules. Distributors must also establish where the rules will be most effective. First, it is important to choose the right level in the product taxonomy as the basis for the cross-selling rule. The most detailed level or SKU level usually does not work well because no pair of items has sufficient support to be applied. The next level up in the product taxonomy which is the model, product, or product line is usually the most appropriate level for defining cross-selling rules.

Second, it is important to choose the right level of unit for analysis. Many cross-selling rules are applied at the level of a transaction. However, other cross-selling rules take a larger unit to analyze, for example all purchases in a one-month period or even over the lifetime of the customer. The choice of which unit to analyze should be determined based on support for the rules. Analyzing support at the transaction level may not provide enough information to identify the relationship or the cross-selling opportunity.

While there can be many different possible cross-selling rules, if is often the case that as few as 20 to 30 rules will be sufficient to provide significant lift. But it is important to consider who will be using the set of rules and in what setting. For field sales, it is important to have a small set of rules that can fit on a single page of paper.

Technology can significantly expand how many rules may be used at any given time. Inside sales people who are supported by a CRM product can act on more rules effectively because the rules are built into the system. The same is true for e-commerce platforms.

Cross-selling is the most relevant technique for growing revenue with existing customers. Analytic techniques are essential for cost-effectively identifying and applying relevant cross-selling rules. The application of even a small number of the right rules can provide sales lift of 1 percent or more.