Up Close And Personal
Understanding your customers' online buying habits is now business critical, so pick your clickstream analysis tools carefully
By Don Nachtwey
In the brick-and-mortar world, the profit-maximizing motive of the merchant and the convenience-maximizing motive of the consumer are out of balance. What results is an optimization process where merchants attempt to satisfy a cross section of consumers, while focusing on their average customers where the greatest volume, and consequently the greatest profit margins, exist.
But few consumers have an average consumption pattern. The ideal experience for consumers would be to walk into a store where the racks have only their size and color, the music store stocks only their favorite artists, or the checkout counter at the computer store already has the correct paper and cartridges for their printer.
But the Internet can potentially move merchants' motives into equilibrium with consumers' motives. Consumers can communicate their demands directly to the merchant. The merchant can in turn communicate the price and availability directly to the consumer. The measurable benefit to the consumer is savings in time and money; the measurable benefit to the merchant includes a significant reduction in inventory and sales-related expenses.
One of the major results of this market-efficiency movement is a power shift from merchant to consumer. Steve Larsen, senior marketing vice president at Net Perceptions Inc., argues that this power shift has created a free-agent consumer than can demand when, where, and what price for products and services. The consumer is only a click away from a new merchant who can better satisfy their demands.
Fortunately for merchants, data flows in both directions. Consumer browsing behavior represents data collection opportunities. This data can become a powerful corporate asset that your company can leverage into greater, more profitable market share.
Implied Communication
E-commerce has expanded from its early days when communication between merchants and customers was limited to email, and credit card numbers were still only provided over the phone. Merchants continue to receive direct communication from consumers, but they are now able to recognize implied communication.
Implied communication is the information that a consumer supplies the merchant through sources such as call centers, banner ads, and the clickstream. These sources are recorded consumer touch points outside the sales transaction that can be used for promotional campaigns, personalized content, or for dynamic targeting of offers. If you have ever shopped for a book at Amazon.com, you know it displays a list of titles other shoppers have purchased along with the book that you are considering. This classic up-sell and cross-sell approach relies on your click-through information.
A growing number of product and service providers are trying to manage consumer-implied communication by capturing and quantifying it into an intelligence infrastructure. This article looks at the importance of the clickstream in an e-commerce intelligent infrastructure and some of the ways companies are leveraging their customer data.
Understanding Clickstream
The clickstream is synchronized communication between your company and your customer. Customers communicate through their client software (usually a browser) to a company's hosting server, which acts as a preprogrammed customer service representative. When a customer requests something, the server relies on its rules engine to deliver an appropriate response.
But in electronic communications, customer queries can seem ambiguous. An automated response to a click on a link may be easy, but a customer search query generates confusing results. In a brick-and-mortar drug store, you can ask a clerk where to find eye care. The clerk can see that you are wearing glasses and carrying an empty bottle of lens solution. Without additional inquiry, the clerk can direct you to the contact lens section where you will find an assortment of lens care solutions or mention a store promotion on your particular brand.
Now if you tried the same query using a company search engine on its Internet store, the server returns a list of products and services that matches eye and care. You are now forced to search through eye drops and eyeliner or generate additional queries before you can find what you are looking for.
The good news is that your search and clicks can create a digital footprint, stored in the company's database, which becomes more detailed each time you visit the site and can predict your future needs. The problem is getting the merchant to recognize your footprint. The clickstream can be a beast made up of log codes, promotion IDs, results errors, and more.
Net Perceptions, Macromedia Inc., and Accrue Software Inc. are among the companies capturing and interpreting customer clickstream behavior to offer merchants both profit maximization and a way to personalize their visitors' experience. Each company defines the clickstream analysis problem in its own way and offers a different approach.
Net Perceptions
The Net Perceptions technology focuses on customer lifetime value. Increasing profit depends on attracting and retaining customers with the greatest expected lifetime value. Customer loyalty is a condition of lifetime value and will become instinctual if merchants:
* Make it easy for customers to find what they need
* Present promotions based on a customer's interest, style, and perceived need
* Provide consistent customer information across all company divisions.
According to Steve Larsen, 98 percent of all Web visitors are browsers. Of the 2 percent that are buyers, 10 percent are first-time customers that never make another purchase. These customers are the worst for a company, because they are usually reference customers that require a kickback to the referring affiliate, or they are responding to some form of loss-leader promotion. They infect the database and waste future marketing dollars, continuing to add to the initial sale loss.
For Net Perceptions, customer analysis alone is insufficient to predict lifetime value. Their solution incorporates products and promotions as well as people. For instance, Garden.com has more than 20,000 products, yet repeat buyers tend to purchase from a selection of 25 products. Understanding this reality and building promotions around these products make it easier for potential repeat customers to become first-time customers. Also, understanding where garden shoppers congregate on the Web lets Garden.com place their promotions in locations that will attract repeat buyers.
There are four components to the Net Perceptions solution:
* Commerce solution. This personalization engine learns more about each customer by dynamically pooling knowledge with relevant data from other customers.
* Intelligence channel. This mining-and-reporting engine provides insight into visitor behavior. (See Figure 1.)
* Knowledge management. This com- ponent allows knowledge to be stored across the enterprise. Knowledge experts in inventory can share information with their counterparts in merchandising or finance and vice versa.
* Core personalization technology. This recommendation engine interacts with customers on a one-to-one basis, automatically learning more about their individual wants, needs, and interests.
FIGURE 1 Net Perceptions' Intelligence channel
The Net Perceptions toolset extends beyond the Web to pricing, merchandising, inventory, and advertising decisions in the physical stores. For example, Net Perceptions determines the best price based on manufacturer availability and product turnover. It also analyzes a company's plan-o-grams (a graphical rendering of a retail product layout by store, aisle, and shelf) to determine where improvements can be made in product placements and the optimum size and location for advertisements.
For the merchant, in addition to customer loyalty, the Net Perceptions solution provides:
* Deeper customer insight
* Lower retention and acquisition costs
* Improved customer service without increased costs
* More cross- and up-sell opportunities for more effective advertising.
Bertelsmann AG is the third largest media company in the world, with a database that contains more than nine million products. Their sites include bol.com, Barnesandnoble.com, several Random House sites, and more. Their typical visitor spends no more than five to 10 minutes on the site. Bertelsmann liked Net Perceptions collaborative filtering engine for reducing customer's product search time.
According to Andrew Dorward, director of personalization at Bertelsmann, they have a three-phase implementation. In Phase I, which is already deployed at bol.com, collaborative filtering is used to make product recommendations based on three dimensions:
* Bought dimension. A customer's purchase history determines the recommendations they receive.
* Like dimension. Visitors specify their interests in a personal section of the site. Collaborative filtering combines the voluntary information with the anonymous data of like-minded customers to generate appropriate promotions.
* Page dimension. The site develops an interaction with the customer that begins a data collection process. Once a trend is recognized, appropriate offers can be recommended.
Phase II will gather and combine all visitor information into a data warehouse where it can be analyzed for clustering and segmentation. Once the segments are built, intelligent decisions can be made on product campaigns. Phase III, building upon the efforts of Phase II, will combine business rules with collaboration filtering to generate dynamic offers based on predictions of estimated buyers under different scenarios.
For Dorward, Net Perceptions' personalization capabilities have a solidifying effect. Personalization fosters loyalty and as data is collected from new and repeat visitors, the program will produce a higher level of accuracy. Dorward likens this process to a tree shedding leaves. With a few leaves spread on the ground, the results are modest. But when all the leaves are touching each other, content can become highly personalized for each visit.
Macromedia E-Business Solution
Macromedia's Aria is a Web measurement and analysis product designed with the idea that Web sites with personalized content create brand awareness, improve customer acquisition, and generate more revenue. Macromedia's personalization approach includes customization, rules-based personalization, and adaptive personalization.
Customization lets visitors change pages on a Web site to fit their needs. Rules-based personalization lets the merchant specify fixed rules to change a site based on visitor behavior. The Amazon. com example given earlier is a rules-based implementation. Suggesting an order of chocolates when receiving an online order for flowers is another example of rules-based decisions.
But rules-based systems don't learn or adapt to user behavior in real time and fail when the site can't easily predict the response to an offer. Because products, customers, and business models change, rules-based systems require constant maintenance.
Adaptive personalization picks up where rules-based personalization leaves off. Adaptive systems can follow a digital footprint and select the right content to present or the appropriate product to recommend. A site with adaptive personalization could observe that buyers of digital cameras also purchase video cards, batteries, and storage systems and automatically recommend them.
Macromedia uses collaborative filtering to create Mentors -- people who have similar behaviors. Macromedia uses Mentors to predict content or product suggestions that might be of interest to the individual Web visitor.
Holly Lugassy, the product manager for Aria, defines the product as a vertical solution that increases its value with each site visit. Aria includes the following components:
* Server Monitor analyzes the Web server's log files and specifically looks at what happens inside a user session, including page navigation, page duration, and click-throughs. This data lets the merchant build a soft-return investment analysis on promotional campaigns.
* Network Monitor tracks the activity coming across the wire such as data packets, IP addresses, image downloads, and broken links. IT departments can use Network Monitor for optimizing the site, but the marketing department can also use it to analyze traffic patterns.
* Application Monitor is an e-commerce monitor that tracks events, such as customer shopping-basket activity, for specific applications by BroadVision Inc. and Art Technology Group (ATG) Inc.
Captured data can be reported dynamically or stored in the data warehouse where it can be accessed using Business Objects' OLAP tools that come with the product.
Another benefit to Aria is the ability to integrate with other offerings from Macromedia. Macromedia also develops and markets the personalization engine Likeminds and is a major publisher of Web development tools, such as Dream Weaver and Flash. Web developers can integrate Dream Weaver objects for Aria that respond to customer site activity by generating personalized content.
Steve Danker, CIO of Musicland Stores Corp., which operates the Sam Goody music retail chain, adopted the Macromedia solution as a way to differentiate the company in the retail entertainment market. The key for Danker was Aria's ability to scale both horizontally and vertically and Likeminds' ability to provide product recommendations using collaborative filtering. (See Figure 2.)
FIGURE 2 Macromedia's collaborative filtering is used at the SamGoody.com site.
Musicland logs more than 100 million transactions per year. Although growing rapidly, Web site transactions account for only a small fraction of total transactions. In addition, Musicland supports a frequent shopper club with more than one million card-carrying members, many of whom are teenagers who don't own a credit card. The Macromedia solution lets Musicland respond with consistent content regardless of whether the customer is visiting the Web site or a kiosk inside the physical store.
Accrue Insight 5
Accrue's focus is to maximize the return on investment (ROI) of Web-based initiatives. According to Vito Salvaggio, vice president of marketing at Accrue, reducing customer-acquisition costs, displaying relevant content, identifying profitable affiliate channels, and improving conversion rates will maximize ROI. (See Figure 3.)
FIGURE 3 Insight 5 report identifying profitable affiliate channels.
Insight tracks and collects data from the user's visit, across multiple servers and applications, for the duration of each visit -- including before, during, and after a transaction. Network-level data provides a rich depiction of each user's visit to a site by recording events such as resets, stopped downloads, and visit duration. Combining this data gives a "unified view" of the customer that can distinguish the most and least profitable customers. You can then tailor promotions, ads, and content to the individual, resulting in improved browse-to-buy conversion and customer retention.
Insight includes the following features:
* Network Collector monitors packets of information as they are transmitted over the network, capturing demographic data, user events, and network activity that the Web server does not log. With this information, you can better evaluate the user's experience and determine if download times are too long, files are too large, or visitor demand is overwhelming Web resources.
* The Analyzer processes raw data from any combination of data collection points into analysis-ready information. The Analyzer weaves a visitor's activity across all the Web servers in the enterprise into a "visit," including information such as time spent per page, reset requests, and visit duration. Analyzer generates a path analysis that shows how the visitor's navigation decisions are distributed across the Web site. So if the desired path for the merchant goes from product promotion to shopping basket, but the consumer's most common path is from promotion to more information to exit, Analyzer will expose it.
* Insight's data warehouse stores all captured information, access, and analysis. Import and export APIs let the user integrate marketing and census information from offline sources into the warehouse, providing a more in-depth understanding of the user. The warehouse supports multidimensional views of customer behavior that expose profitable sweet spots that can be the basis of future promotions.
More than a year ago, Autozone Inc. was looking for a way to monitor its site performance. It chose Insight 5 for its packet-sniffing capabilities and reporting tools, mainly to ensure that its Web site did not have broken links or problems rendering pages in certain browsers. According to David Wittenberg, an e-commerce analyst with Autozone, the role of Insight 5 has since quickly expanded across three areas: monitoring operational server activity, tracking content and development problems, and judging the effectiveness of marketing and advertising.
Insight also has a feature that can convert the access logs to an Accrue format. From there, the data can be imported into Excel or another analysis tool. Insight includes a thin-client reporting tool that Autozone analysts can access from their Web browser. Insight also provides native access to Informix Corp.'s Red Brick RDBMS, Autozone's data warehousing platform.
Ahead of Its Time
The supply of solution providers has grown considerably. I've discussed products from Net Perceptions, Macromedia, and Accrue. But many more take a different bite at the problem. BroadVision and Epiphany Inc., for example, offer clickstream tools as part of a Web-management solution that includes transaction and presentation components. Others such as DataSage Inc. (recently acquired by Vignette Corp.) focus on credit scoring, and Personify Inc. builds tools for creating customer segments.
Although suppliers are plentiful, demand for these tools has yet to reach the tipping point. A recent survey shows that the primary interest of most e-tail sites is to provide product information. Only 36 percent of e-tailors identify online selling as their primary goal, while 28 percent identify brand loyalty as the primary goal. Some 82 percent of those surveyed measure Web site success by the number of hits to the site -- the "Wow" factor (Wow, we are getting a lot of page hits). Only 12 percent measure success by the look-to-buy ratio.
It may be that these tools are slightly ahead of their time. E-tailers may not be ready to invest $100,000 or more on personalizing electronic stores that only generate a small percentage of their company sales. However, as online shopping becomes more popular and online customers become more sophisticated, personalization will become expected. The stores that guarantee their own future will be those that increase their customers' lifetime value by accurately responding to a customer's needs and styles.
Don Nachtwey (dnachtwey@yahoo.com) is a product manager for e-commerce.
|
