Retailers work on creating enticing cross-sells online
By Linda Punch
Effective cross-selling and upselling is a mixture of art and science—coming up with that certain item a customer just can’t refuse to buy. But most online merchants have yet to master the craft. Although 76% of online retailers use at least one method of cross-selling or upselling, only 9% of online buyers say they used site suggestions, according to Jupiter Research.
“Retailers are trying their best but it’s hard to get a lot of traction out of cross-sells and upsells,” says Patti Freeman Evans, Jupiter Research analyst. “Consumers are not making impulse buys.”
Getting easier
But that task may become easier with the increasing use of analytics software. These new automated programs are helping retailers refine the cross-selling/upselling process, leading to additional sales. That’s the case at Sierra Trading Post, which began using an automated cross-selling and upselling program from web analytics company Offermatica Corp. a little over a year ago.
The Offermatica product replaced a manual system that used “a best guess method” on select items that produced mixed results, says Chris Lange, web operations manager. And while there were some initial problems with the automated process that required fine-tuning, the results were gratifying: An increase in average order value of 2% to 5% and “sometimes greater,” Lange says.
Online jewelry retailer Ross-Simons has had similar results with an automated program from Ecometry Corp. “We do in the neighborhood of $1 million a year in upsells so it’s a pretty nice program,” says Terry Matthies, vice president of information technology at Ross-Simons. Ross-Simons had $63 million in Internet sales in 2005.
Automated programs use a variety of methods to generate cross-selling and upselling recommendations, and are being integrated into e-commerce platforms, web analytics programs, site search, and other types of software commonly used by online retailers. Some vendors, such as Ecometry, also offer stand-alone cross-selling and upselling modules.
Recommendations are based on a broad range of information, including a customer’s past purchases as well as overall customer purchase patterns. Some programs track product performance, including which items customers most often buy together, or develop recommendations based on the combination of items that an individual places in the shopping cart.
And in some cases, automated cross-sell and upsell programs will take into account click-stream behavior, customer reviews, and profit margins of items.
Dual approach
Ecometry takes two approaches to cross-selling and upselling, says Brian Dean, vice president of marketing and strategy. The first is a merchandise-driven strategy, in which the retailer selects the cross-sell and upsell candidates for each item. The retailer can configure the system to make substitute suggestions in case an item sells out.
At Ross-Simons, which uses Ecometry’s commerce suite, the retail staff identifies upsell candidates at the web site and call center based on factors such as pricing and quantities available, Matthies says. The information is entered into the system by the retailer’s operational staff.
The second approach is analytics-based—recommendations are based on what previous customers with similar shopping patterns purchased. “It’s similar to what you find on Amazon, which says ‘people who purchased this also liked this,’” Dean says.
Ecometry’s software builds “relative strength indexes” on possible cross-sell and upsell candidates based on the items a customer puts in the shopping cart, and updates the index each time the customer adds an item. That data is used in tandem with information on customer purchasing patterns and product preferences to produce recommendations.
Ecometry’s system uses a common engine for the web, retail store and call center and stores information on a customer’s purchases in all three channels in a central data warehouse. “As you add things to the cart, it’s building a composite affinity calculation for different product categories and individual products,” Dean says. “When you’ve got three items in there, it has a different relative strength index for a product category or a product than if you just had one.”
For example, if a customer puts two country music CDs and a classical CD in a shopping cart, the system most likely will recommend another country album, he says.
Filtering
The Ecometry system allows retailers to give top priority to the merchandise approach to recommendations, Dean says. “If you don’t have a merchandise rule, than you go to relative strength indexes,” he says.
Ecometry’s system also filters out hugely popular items that could skew results, Dean says. For example, a newly released DVD might show up in shopping carts with a large variety of totally unrelated items.
The system also filters out frequently purchased items, such as batteries. “You don’t want to recommend batteries all the time even though batteries are purchased with a whole host of things,” Dean says. “There are certain things you filter out just because you don’t need to tell people about those things.”
Offermatica’s system uses merchant input in setting up the system, but then uses its Lazy Loading technology to automatically build a database, says Erin Casey, marketing manager. With Lazy Loading, product information automatically is sent to the database at the same time an order confirmation page is sent to the customer.
“As people purchase things, the product information is passed back to our system,” Casey says. “It makes it very easy for the merchant. They don’t have to upload information—the database is automatically being built for them.”
Automated systems don’t necessarily lock retailers into one format. The Offermatica system can make recommendations in categories such as best seller, best seller in category, most viewed or in whatever fashion the merchant chooses. As with Ecometry, merchants can designate which products get top priority. For example, a retailer might want to push a seasonal item to the top of the list. “The merchant is very important in this process,” Casey says.
With the Ecometry system, retailers also can set conditions for cross-sells and upsells. “Some people want to use product recommendations to move overstock—it’s a way of avoiding markdowns,” Dean says. “Other people want to go with a static list—they want to put items on sale for a week.”
Measure results
But for retailers to be able to select the best strategy, they must be able to measure results. Many times, what might seem like a slam-dunk cross-sell idea flops. “People think just because they implement cross-sells, it’s going to drive up average order value and sales, but that’s not necessarily the case,” Casey says.
That’s what Sierra Trading Post found when it began testing which minimum price levels worked best for cross-selling, Lange says. In the initial test, the levels were $10, $20 and $30. “We quickly abandoned it because $20 and $30 were doing much better than the $10 test,” he says.
Sierra then tested $20, $30 and $40 minimums and found that $20 and $30 levels worked while the $40 level was “outside our customers’ comfort level,” Lange says.
Offermatica’s program enables retailers to do A/B split tests or multi-variate tests on cross-sell and upsell strategies, for example, whether to make recommendations on the product page or the checkout page. Offermatica has found that inexpensive related items, such as socks or belt, work best on category pages, while more expensive coordinated items are more likely to sell on the checkout page.
A recent study from the E-Tailing Group Inc. found that 88% of retailers have cross-sells and upsells on the product page while 62% put recommended products on the shopping cart page.
Ecometry is enhancing its predictive response model to monitor the success of product recommendations and then modify its behavior to push the most frequently accepted recommendations to the top, Dean says. The software checks whether the recommendation was accepted or abandoned. It also tracks the profitability of upsells.
Ease of use
While the technology used in automated cross-sell and upsell programs is sophisticated, retailers usually find it easy to use. At Sierra Trading Post, the program can be modified by changing four or five lines of code per page, Lange says.
And Matthies says training a new retail staff member how to use the Ecometry program requires only about 1.5 hours. “You don’t need to be a technical genius to do it,” he says.
But while the idea of automated cross-selling or up-selling is appealing to retailers, many simply can’t afford the technology, says Lauren Freedman, president of the E-Tailing Group Inc. “It’s a resource issue,” she says.
Other retailers may be afraid to automate the process because they fear product selection won’t be as relevant, Freedman says.
There’s some basis for that concern. In execution, automated cross-selling can fall far short of expectations, and, in some cases, have disastrous results. Witness the experiences of Walmart.com and Amazon.com. At Wal-Mart, an automated system linked African-American themed DVDs with the DVD of “Planet of the Apes,” while Amazon’s automated system suggested books on adoption to customers seeking books about abortion. Both incidents generated a flurry of bad publicity for the retailers.
But those cases are rare and extreme. “That kind of problem crops up when you’re trying to associate contextual search to a product as opposed to a product-to-product affinity,” Dean of Ecometry says. “What we’re returning is a pure relationship that these things are purchased together.”
In theory, an automated system based on a product-to-product relationship could produce the “convenience store scenario where they say beer and diapers are purchased together frequently because husbands are sent out to get diapers and they pick up beer,” Dean says. “There really isn’t the risk that the engine is going to interpret search criteria and tie products to that. Your only risk is that your recommendations on occasion may seem odd.”
And such problems can be avoided if retailers are careful about the rules they write for the system, Freeman Evans says.
Where it makes sense
What’s more, some items simply don’t lend themselves to cross-selling or upselling. Although it’s easy to cross-sell CDs or DVDs, that’s not the case with apparel and footwear, says Freeman Evans.
“If you like Sting, you’re going to want all of his albums,” she says. “But if you’re into Kenneth Cole, it doesn’t necessarily mean you’re going to like his next collection. It’s very taste-oriented and it makes it more difficult to get the right product.”
Another factor that diminishes the effectiveness of automated cross-selling or upselling is that customers often don’t know what specific item they’re looking for when they get to a site, Freeman Evans says. “They might be looking for jeans, but which jeans?” she says. “If I don’t even know what jeans I’m going to buy, how am I going to decide on what top or shoes or belt I’m going to buy?”
Despite these difficulties, cross-selling and upselling are established practices in the offline retailing industry and retailers will surely find a way to do it online as well. “It’s second nature to us,” Matthies says. l
linda@verticalwebmedia.com
How to close those cross-sells and upsells
• Consistently position cross-sells and upsells on every product page
• Recommend an average of 3 items to complement each product
• Be sure related items are relevant to each product
• Try innovative category-centric bundling techniques enabled by Internet technology, such as shop by room/by outfit
• Present recommendations to catch customers’ attention, such as “Our experts suggest…” or “Others who have purchased this have also bought…”
• Use promotional offers to entice; for example, limited time discounts on related items
• Include services, such as extended warranties, as well as products
• Use the Internet’s flexibility to test offers, changing them frequently onsite and across channels
