|
Last year was a breakout one for e-commerce.Many consumers opted for the ease of the click-and-order stores instead of the annual grind inside brick-and-mortar aisles. The aggressive sales goals that analysts set were blown away by consumers eager to adopt the electronic buying medium.
Still, 1999 had its disappointments. The demand for products distributed by e-commerce stores exceeded supply, resulting in a last-minute rush to the local mall to grab something to put under the tree. So will the fulfillment problems of 1999 translate into lowered expectations for 2000? Not likely. The benefits of online shopping are too great, and consumers intuitively understand that the demand in this country’s economy rarely exceeds supply.
I expect two reactions to the fulfillment problems of 1999: First, shoppers will plan ahead. Buyers whose November 1999 orders were unfilled will make their 2000 purchases in October or September. Second, e-commerce stores that promoted products that were out of stock in 1999 will lose those customers who tried to buy those products. Promoting and selling a product on the Web and then not delivering the product is just like breaking a promise. You promised the customers convenience, savings, and personalized service. If you fail to deliver any single one of them, you could lose that customer forever. In the electronic marketplace, trust is an essential part of customer loyalty.
Market Efficiency
So how could 1999’s downside have been avoided? The answer is: with more accurate information. The market medium that the Internet represents is an evolution toward perfectly efficient markets, or what some have called reverse markets. Economists define a perfectly efficient market as one in which a product’s price is set at a point where the consumer’s demand schedule intersects with the producer’s supply schedule such that market demand is satisfied and market supply is consumed.
The key ingredient to a perfectly efficient market is perfect information. Perfect information implies knowledge of all that has occurred as well as all that will occur. Such perfection is impossible, so the most we can hope for is near-perfect markets. The stock market is such a market; a security price is set when agreement is reached between the buyer and seller. The price that investors are willing to pay is based on a valuation formula that considers past performance, expected performance, future market conditions, and so on.
Evidence of near-perfect markets is springing up across the Internet. Priceline.com, for example, creates a clearinghouse for capacity-based markets such as airlines and hotels. Consumers offer information on their willingness to pay for a seat or a room, and sellers offer the price at which they can make that seat or room available and still make a normal profit. Another example is Mercata Inc., which lowers the price based on the number of bidders for a particular item.
Near-perfect markets in which the consumer can offer a price are the easiest to create using the medium of the Internet. Most consumers go through a budgetary analysis for luxury items such as vacation travel. They know to the nearest dollar what it would cost to spend their vacation at Aunt Midge’s house. So they can make an online offer for a Florida vacation at a price equal to the cost of a vacation with Aunt Midge or add a few dollars to see if there is a seller for a Hawaiian vacation. If there are no sellers, just remind the kids how much they love playing badminton in Aunt Midge’s backyard.
But for most goods, information doesn’t exist because consumers do not have an explicit price in mind they would be willing to pay. Typically, they make needs-based decisions on seasonal or style changes or depletion of household goods. For these goods, the consumer is a price taker, not a price maker.
Pricing in the Electronic Marketplace
In these markets the seller may set the price, but the advantage does not go to the seller. Although consumers may not have an explicit price in mind, they do have an implicit price they would be willing to pay. For instance, if Mrs. Smith can’t find a coat for her son Billy for less than $100 to replace his old coat, she may make the old coat last another year. The merchant will then have to offer the new coat at a 50-percent clearance price at the end of the season. The costs to the merchant of mispricing the coat are lower total revenue and revenue that goes unrecorded until later in the year.
However, with Web-based information, merchants could’ve predicted what Mrs. Smith was willing to pay for the coat and set the price that would generate a profit (and maybe sell a pair of gloves, as well).
Consider the following hypothetical: Mrs. Smith clicks over to Coats.com from a Yahoo search for discount clothing and starts to browse through the different styles and sizes. She selects two or three coats for which she requests more information. Finally, she focuses on the one she likes best. She may even go as far as placing the coat in a shopping cart. Then she abruptly leaves the site to browse Winterwear.com
Coats.com has captured the entire session, offering key insights into this consumer’s buying behavior. The problem is that Mrs. Smith kept browsing without buying. Let’s stop here and discuss the session information for Mrs. Smith’s visit before taking a look at what Coats.com needed to close the sale.
Web-Based Data Sources
When Mrs. Smith visits Coats.com, she is essentially requesting data from the Coats. com server. Each request is a transaction that generates a record in a log file. The log files that are usually available include:
•Transfer Log—where every transaction between the server and browser is recorded with date and time
•Host Field—shows the host server making the request
•Authuser Name—contains the authenticated user name that a visitor needs to gain access to a password-protected site
•Time Stamp—contains the date and time
•HTTP Request—contains the get call that a visiting browser is requesting
•Status code fields—include Success, Redirect, Failure, and Server Error
•Error Log—populates when a transaction on the Web site results in an error
•Referred Log—contains the URL from which the Web site visitors originated; possibly a link from another page or a search-engine hit
•Agent Log—records the name and version number of the browser making the request.
Inside your page, the cookie file can track the navigation path of your customer. Cookies are a technology Netscape designed to let merchants create the impression of a continuous shopping experience over several unrelated user sessions. They are small text files that a Web server passes to the client computers, where they are saved so the server can identify users and track their navigation path when they return. Cookies are what let a commerce site welcome you by name.
Another source of information is generated through a CGI script on the merchant’s Web site. Visitors generate this data when they enter information in online forms. These forms usually collect information about user occupation, profession, ZIP code, gender, and age.
The Power of Web Information
Now let’s get back to Coats.com It collected a lot of information about Mrs. Smith’s visit. How can it use the information to make a sale? In truth, it’s probably too late to make that sale. She has probably found a coat at Winterwear.com or some other store. Nor will the data help to make future sales because the coat is likely to last several years.
In the world of e-commerce, it helps to know something about each of your customers before they visit your site. If the data generated from the session were to combine with the consumer’s profile information, you could generate targeted offers dynamically. For example, let’s say Billy’s family is from Minnesota. If Mrs. Smith carries an electronic profile (perhaps in the form of a cookie) with her to the Coats.com Web site, when she clicks on juniors coats, a special offer would appear for a fleece lined windbreaker effective to {30o F—which happens to be the record low for her ZIP code. In addition, if she buys the coat, she can get 50 percent off a pair of Gore-Tex gloves.
Infomediaries are especially effective at collecting and providing profile information. Infomediaries are businesses that capture, manage, and maximize the value of consumer information. (My company, e-centives.com, is an example of an infomediary; others include Mypoints.com and Cybergold.com ) These companies register customers with the promise that they will receive only those product offers that are consistent with their interests and needs. To do so, the infomediary collects demographic and interest information on each of its subscribers in the registration process. At the same time, the member gives permission to the merchant partners of these infomediaries to advise them of relevant product offers.
If Coats.com is a merchant partner of an infomediary, then when Mrs. Smith visits the site, the company may receive information such as: the customer is female, she is from Minnesota, she has two children, her annual income is between $50,000 and $75,000, and she likes to travel. Now Coats.com can tie this information into the data it may have built in its data warehouse or into a direct marketing database that tracks the buying behavior of customers that match Mrs. Smith’s profile.
Modeling Consumer Behavior
When the data has been formatted, it will only add value if it supports business decisions. In the pre-Web world, the datacaptured by Coats.com captures would be stored for use by analysts, who would break it down into a story of “what happened,” and subsequently by planners, who would develop strategies for the next business cycle. In the world of e-commerce, trade is dynamic, marketing is instantaneous, and business corrections are made in real time.
Let’s use the game of baseball as an analogy. In baseball, it is not uncommon for a rookie to get the hit that wins the game. The probability of such an occurrence is not that remote. Let’s say the game is tied in the bottom of the ninth, the bases are loaded, and the count stands at three balls and two strikes with two outs. The rookie gets the game-winning hit on a fastball. The rookie has no experience in that situation, yet he is prepared for the fastball. He expects the fastball because his team has been in that situation many times during the year, and the pitchers have thrown a fastball nine out of 10 times regardless of the batter. Across the league, 90 percent of the pitchers have thrown a fastball in the same situation. And throughout 100 years of baseball, pitchers have always thrown the fastball. Thus, the pitcher has the real disadvantage, not the rookie. He has never faced the batter before, he doesn’t know if the batter is better at the high fastball or low, inside fastball or outside.
In e-commerce, the merchant plays the role of the pitcher. When new customers visit the merchant site, the merchant has no information on them. It doesn’t know their names, let alone their merchandise likes and dislikes. In baseball, the good pitcher will use cues and sensors to turn data into knowledge and knowledge into decisions. The pitcher will take inventory of his own strengths: His best pitch is the high fastball; his strongest defenders are on the left side, his best center fielder is on the bench with an injury. Then he will take cues from the batter. Is he left-handed or right-handed? Does he stand close to or far from the plate, forward or back of the plate? In our example, the pitcher also has five pitches’ worth of data (three balls and two strikes) to consider. Did the batter swing on strike one or strike two? Did he lay off the high pitches and swing at the low pitches? He has a total of 30 seconds to process the data and determine his pitch.
The electronic merchant has similar cues. The merchant knows its most popular products; it also can see which “door” the customer enters through. (If it’s Lycos, for example, the customer is probably under age 35.) Customers may be referred by an affiliate that caters to certain demographics or interest communities. They may move directly to one department and spend more time viewing one product vs. another. This data may be all that is available to the merchant, but it still should be enough to derive a measure of buying preference. The difference between baseball and e-commerce is that the merchant has only has about five seconds to process the data and determine which product to “pitch.” If the customer has to wait more than five seconds, the merchant will also have an exit code to add to the data file.
In the case of Coats.com, the merchant must be able to predict what Mrs. Smith will do when she arrives at the Web site. Decision tools must be in place that look at all the available attributes in the data. They then must be able to model it and formulate a “score” that communicates how she will behave, how she will navigate the site, and the probability that she will execute a sale when exposed to a certain promotion.
Now, what if Mrs. Smith is responding to a promotion that was placed on an infomediary site? While Coats.com is expecting the promotion to generate a modest amount of traffic, the infomediary has registered a million new members from its own promotion. Suddenly, Coats.com traffic has spiked and the promotion has more than quadrupled expectations. Will it be able to fulfill the demand? The transaction system is tied to accounts receivable, and revenue is booked. But will the data flow into the inventory system such that the product will be available? Will fulfillment have sufficient packaging? And will distribution be alerted to the spike in demand and manage the requests so that one or two centers don’t get most of the load?
A similar situation occurred to one of our merchant partners. The promotion exceeded expectations and orders kept coming in after inventory was exhausted. The data was available, and were it properly managed, it would have alerted production managers to prepare for the spike in demand—or at the very least, warn the merchant to halt the promotion until inventory could be replenished.
Moving Forward
If the data—as well as tools from companies such as HNC-Aptex Software Inc., DoubleClick Inc., Epiphany Inc., and Net Perceptions Inc.—were available in 1999, why were merchants unable to meet the consumer demands of the holiday season? Failure to adopt decision-support technologies may be partially to blame. The data is being collected. The demand for disk storage is increasing 130 percent annually while human population growth is at 3.1 percent. But only 2 percent of the data being collected is used. It may be that IT managers have a negative attitude about the process, considering Web content too unstable, transformation of hypertext into a structured database too difficult, or the data of little value because it’s external to the business.
Another problem is maturity. Traditional merchants understand how to collect and use point-of-sale data and direct-marketing data. On the Web, data arrives in the form of log files, cookies, CGI sessions, and so on. This data must be normalized in order to integrate with the schema that exists in the data warehouse, a process that has yet to enter the mainstream.
Maturity was also a problem for decision-support technologies. These tools are in their first or second release and were put through their paces only in 1999. In addition, these tools were working with limited data; the modeling necessary to predict customer behavior requires a learning process that is based on past customer behavior. More customer transactions will result in more patterns of behavior, which will create smarter predictive models.
When these models develop an 80-percent accuracy rate, the results will flow freely through the supply chain, tuning systems to offer the right content and products that are most likely to be accepted, allowing manufacturing capacity to synchronize with customer buying patterns, and optimizing distribution systems.
The speed of the Internet and the growing competition in the tool markets should rapidly result in product maturity that will be evident when consumers hit the electronic malls at the end of 2000.
Don Nachtwey(don@ecentives.com) is product manager for e-centives’ Campaign Manager service. His career in IT spans more than 10 years, and he has been working on Internet commerce products since 1994. The opinions expressed here are his alone and do not necessarily represent those of e-centives.
|
|
|
|
|