A Matter of Metrics: Using Web Data to Improve Sales Performance

For Internet-based retailers, it is no longer enough just to flash hockey-stick sales curves at investors and impress them with stratospheric sales figures each quarter. What they require are appropriate benchmarks, which managers and investors alike can use to see how well an online merchant is really doing.

Decomposing sales and buyer data to gain a better understanding of the way visitors behave is even more important on the Internet than in the bricks-and-mortar world. According to Forrester Research, a consulting firm in Boston, more than 70% of Internet retailers in 1999 had a conversion rate of less than 2%, which is to say that of every 100 visits to a retail website, only two resulted in purchases. Industry experts suggest that market leaders such as Amazon.com have conversion rates no higher than 15% to 20%.

Even this data is fuzzy at best, however. While firms tend to measure conversion rates as the number of people who completed a transaction during a particular period as a proportion of total visitors during that time span, investors and managers would gain greater insights if they knew the probability of each consumer making a purchase during each visit.

So far, the disturbingly low conversion rate among e-tailers has been masked by a steady influx of new Internet users. But this situation is rapidly changing as in the U.S., at least, the market matures and universal access to the Internet comes closer to becoming a reality.

"Saying something like, ‘Sales are up 150% over last year’ means nothing unless we have a strong sense of what sales would have been under some sensible scenario," says Peter Fader, associate professor of marketing at Wharton. Fader, who has been studying grocery-store scanner data for for nearly 20 years, is now investigating web-based data and fitting it into similar marketing models. In June he will teach a course about "Web Metrics: Making the Most of Your E-Commerce Data," which draws upon his research. The course aims at improving the use that managers make of Internet data to help them strengthen their marketing efforts.

Fader’s research shows that the nature of benchmarks and models is surprisingly similar in the online and bricks-and-mortar worlds. Even though the actual psychological processes that people go through are different, their behavioral patterns are strikingly similar.

This has important implications not only for investors who buy shares in these firms but also for managers who devise sales strategies. First, the hockey-stick sales curves reported by most dot-com firms might not be sustainable over time as the influx of new users slows. And if stock prices are just discounted future cash flows, then can stock market valuations which price in expected exponential growth rates be sustained over the long-term?

"Investors need to establish benchmarks about their firm’s performance. One important factor is to figure out how often new customers in a particular period come back to a site to buy again. But that information is very hard to find," says Fader.

For managers, it is important to break up overall sales figures into their component parts, says Fader. Managers should be looking at the amount of traffic that comes to a web site; the rate at which visitors are converted into buyers; the frequency with which buyers return for repeat purchases; and whether buyers spend more when they come back.

"Understanding and forecasting each of these pieces separately and then pulling the pieces back together would help managers get a complete picture," says Fader. "Most firms are doing a very poor job. They do not decompose their figures in the first place, and they also do not perform these separate analyses, if at all."

Moreover, when managers diagnose the results, they also need explicitly to acknowledge that there are tremendous differences among people. "In many cases, it is not enough to say, ‘Our conversion rate is 10%,’" Fader explains. "Some people would infer that ALL the visitors buy 10% of the time. But it might instead be true that 10% of the people buy on every visit, but 90% never buy. Both stories (and many others) lead to the same 10% number, but they have very different implications for allocating resources."

Fader maintains that all metrics are not created equal. Simple summary statistics enamor managers, though they are sometimes misleading or downright meaningless. Example: The statistic that attempts to answer the question, "How many of our current purchases come from past buyers?" In Amazon.com’s case, management is proud to point out that repeat purchases comprise about 70% of overall sales volume. But is a higher number necessarily better? "Consider a leading brand of toothpaste," says Fader. "Virtually all of its sales are from repeat buyers. Does this make it a better business proposition?"

"Lots of e-tailers boast about sales to past buyers, but it has no particular meaning", argues Fader. A better statistic, but surprisingly rare in the online world, would attempt to answer the question, "Of all the people who first bought at this site this month, how many came back and bought again in the next quarter?"

"At first glance both questions might seem similar, but on closer inspection, they are entirely unrelated to each other," Fader explains. "In fact, one could be going up while the other is going down."

Lack of focused analysis also means that it is difficult to understand whether a consumer’s relationship with an e-tailer is long-term (loyal) or just a series of one-night stands. While many web consumers do appear to be loyal—for example, recent research by Fader and other colleagues associated with the Wharton Forum on Electronic Commerce shows that they tend to reduce the number of sites searched over time—much of their apparent loyalty is simply due to the fact that people have very limited activity in any one category.

"If a consumer only goes shopping one time, he’ll appear to be loyal to the brand. Instead, managers need to go deeper, below the limited, sketchy observable behavior that is available for any one consumer. They need to try to understand the latent (unobservable) propensities that people have to be active in a category and to visit or buy at a certain site," Fader explains. "The scary thing is that some sites are telling crazy stories about consumer loyalty, and spending crazy dollars trying to achieve it. In the end, it’s like herding ants. You can make it very attractive for them to stay in one place, but it’s not necessarily because that one spot enamors them. Once the tasty stuff is gone, they’ll scatter again. Some will stick around, but not nearly as many as people would have expected based on the initial observations of the ants’ behavior," Fader notes.

All this means devising new advertising strategies. While a common strategy among several Internet retailers has been to spend heavily on advertising to attract new users, Fader points out that it remains unclear whether such a strategy produces loyal customers or higher conversion rates. Instead, he suggests that it may be time to decompose sales into new and old buyers and then make more informed decisions about whether to use advertising dollars to target repeat customers, versus trying to convert light buyers into heavier users.

Since roughly 20% of buyers bring in 80% of the total revenues of many dot-com firms, these businesses will have to devise strategies to bring back the big spenders by giving them incentives—comparable to frequent-flyer miles or reward points—to shop at their sites. Indeed, e-tailers need to strike the right balance between bringing in new people versus tracking the old ones, especially as the influx of new users slows.

Fader explains that marketing clues for e-tailers may lie in the operations of bricks-and-mortar retailers. Both worlds need to manage promotions and resolve resource-allocation issues in a somewhat similar manner. As the number of new Internet users slows to a trickle, what will keep dot-coms from bankruptcy is their ability to bring back the same customers again and again. Lessons on how to make that happen–and gloomy stories about those who failed to do so–can be found in bricks-and-mortar stores everywhere.

Citing Knowledge@Wharton

Close


For Personal use:

Please use the following citations to quote for personal use:

MLA

"A Matter of Metrics: Using Web Data to Improve Sales Performance." Knowledge@Wharton. The Wharton School, University of Pennsylvania, 29 March, 2000. Web. 27 October, 2020 <https://knowledge.wharton.upenn.edu/article/a-matter-of-metrics-using-web-data-to-improve-sales-performance/>

APA

A Matter of Metrics: Using Web Data to Improve Sales Performance. Knowledge@Wharton (2000, March 29). Retrieved from https://knowledge.wharton.upenn.edu/article/a-matter-of-metrics-using-web-data-to-improve-sales-performance/

Chicago

"A Matter of Metrics: Using Web Data to Improve Sales Performance" Knowledge@Wharton, March 29, 2000,
accessed October 27, 2020. https://knowledge.wharton.upenn.edu/article/a-matter-of-metrics-using-web-data-to-improve-sales-performance/


For Educational/Business use:

Please contact us for repurposing articles, podcasts, or videos using our content licensing contact form.