by Robert Richardson
In the world of traditional sales, few tactics are as common as cross-selling and up-selling. Even customers that have never heard these terms before know what they are. Cross-selling can be as simple as asking, “Need some batteries to go with that gizmo you’re buying?” Or even: “You want fries with that?” Up-selling, of course, takes place when they offer to increase the original purchase by “super-sizing” those fries.
Perhaps the easiest aspect of selling up and across is that you’re dealing with existing customers. They’ve already agreed to do business, and now you just need to figure out somehow the next thing they’ll want to buy. In a conventional retail situation, sellers know these things because they know a lot of your customers personally. On the Web, the same rules apply, but instead of knowing customers by their names and faces, businesses “know” them as transaction logs and “clickstreams.” But put together the knowledge of what’s being sold (“you need special lightbulbs with this lamp”) and the customer data (“people who like this lamp like stainless steel tables and chairs”), and it’s possible to do a good job of creative cross- and up-selling.
NOT AS EASY AS IT SOUNDS
Of course, you first have to make the initial sale, then go ahead and tempt them with extra desirable items. For Web businesses, that part is relatively straightforward once a customer arrives on the site. Simply add the “right” offer to the very next page.
But even this simple act of making the extra offer requires custom scripting and, usually, a database running behind those Web pages. The database is what flips the next desired item onto the page based on what item was purchased first. (See “Your Site: Getting Personal,” May 2000). For now, we’ll focus on figuring out how to choose the right “extra” to offer, based on what’s known about a customer.
Sometimes the right offer is obvious (like required batteries), but sometimes a good offer requires some study and insight. For this part of the equation, you need to engage in what’s becoming an emerging technical field in its own right “data mining.”
DIGGING FOR DATA
Perhaps the first thing to realize, according to Dick DeVeaux, data-mining consultant and professor of statistics at Williams College, is that data mining isn’t actually about matching individual customers to their individual desires. It’s not peering into one buyer’s soul so much as trying to learn the likelihood that this particular customer is going to behave like other customers. It’s about looking at aggregate data and using it to increase the overall performance of offers.
The types of data collected when “mining” a Web site, DeVeaux says, include transaction data (the items sold to each customer), data that users may share from filling in any forms posted on the site, and data purchased from other sources.
For people who aren’t career statisticians, DeVeaux recommends looking through this data by using simple statistical and graphical software with an eye for insights that are “surprising, real, and useful.”
“There can be very substantial payoffs,” says DeVeaux, “but bear in mind that trying to do anything much more sophisticated than that today may cost a small business more than it’s worth.”
DEMOGRAPHIC TOOLS
Although DeVeaux is correct in saying that few data-mining tools are appropriate for small businesses, there are products that provide some initial capabilities in data mining for cross-selling. Smith-Gardner’s Predictive Response makes cross-sell recommendations as the buyers place items in their shopping carts, custom-tailors information pages as the user moves through the site, and flags special offers for customers when they arrive on the home page.
However, Predictive Response requires that the entire e-commerce operation run on Ecometry software. An Ecometry system for 20 users costs $100,000, and pricing for the Predictive Response add-on starts at $20,000.
While that is by no means inexpensive for most small businesses, it can pay off. Smith-Gardner customer ClubMac (www.clubmac.com), an online retailer of computer products, reports a 20-percent increase in the number of line items per order since installing Predictive Response.
“It’s very expensive for us to attract new customers to our Web site,” says ClubMac president Mike McNeill. “For us, the most cost effective way to expand sales is to sell more to our current customer base rather than attract new customers.”
“The system is basically helping us increase profits while servicing our customers better, and thus strengthening our customer relationships,” adds McNeill. “We have seen tremendous results in the short period of time since we employed the module, and know that as our database of information grows, our results will as well.”
DEMOGRAPHICS INTO DOLLARS
It may also be possible to eliminate the tedium of trying to make smart predictions from your own data by using someone else’s. A new service by Austin, Tex.-based startup HotData Inc. delivers business profiles, consumer demographics, and address verification from multiple data providers like Dun and Bradstreet and Polk, over the Internet in real time.
The Web-based e-Luminate service takes information submitted from forms filled in by site users in an encrypted request across the Internet, and returns matching demographics back a second or two later.
The data that’s returned is not specific to the person filling in the form. Rather, it’s data that averages characteristics about a very small geographic area around that person’s address. You can get information about the percentage likelihood that this person is a homeowner, likes golf, or does their own home improvement work, for example.
Having this type of general information may provide some insight as to what a customer’s next need may be. “Those real world statistics are better predictors of a person’s behavior than Web clickstream or transaction histories,” says Ellis Oglesby, HotData’s marketing director.
Still, using HotData involves an up-front fee of $10,000, plus a per-transaction fee ranging from 10 cents to one dollar, depending on the type and amount of information returned.
A final precaution from Dick DeVeaux, however: “You have to keep asking yourself whether knowing a given thing about a customer is really going to be useful to you, compared to the cost of getting that information in the first place. There are big gains to be had, but you’ve always got to be a bit skeptical.”
The least any business can hope for is a better understanding of its customer’s purchasing decisions.
