Be at KMWorld 2018 in Washington DC. Get the Super Early Bird discount when you register with code KMSUPER.

BI branches out

This article appears in the issue November/December 2004 [Volume 13, Issue 10]


   Bookmark and Share

<I><B>More dimensions, more value</I></B>

Business intelligence (BI) is a critical tool in today's competitive environment, but typically has been based on analyses of quantitative, structured data. New approaches that incorporate different types of data are increasing the value of those analyses. Map data, clickstream data and text (see sidebar) all add up to richer, more informative business intelligence.

Mapping a course for growth

The third-largest cable provider in the United States, Cox Communications is working hard to optimize its expansion into new markets. The company offers high-speed Internet service, voice and data communications, and digital telephone services. Because those services all are provided through fiber and coaxial cable networks, the location of potential customers is an important factor in prioritizing marketing initiatives. But to be most useful, location data needs to be integrated with other information such as detailed business demographic data sourced from third parties, customer data, including existing services in place, and scores predicting future behavior such as propensity to buy various services--all presentable in a visual as well as in a numerical format.

For a number of years, Cox Business Services, the business-to-business unit within Cox Communications, had used MapInfo to extract spatial information that was then manually loaded into a database containing third-party demographic data, billing records and other information. The company used a combination of desktop databases, MapInfo Professional and links to various statistical tools as the marketing database.

"We realized that as our company grew, we would need a more efficient and scalable business intelligence system," says Mark Snow, director of marketing analysis for Cox. "After a short stint with a different third-party vendor, we went through an extensive evaluation process with analysts and integrators, and selected MicroStrategy to integrate with MapInfo."

MicroStrategy had been integrated with MapInfo for other customers via an API, but the proposed application was more ambitious. For one thing, Cox wanted a bi-directional capability. Not only should the application be able to generate a visual representation of information in the database, but users should also be able to begin with the map and work their way back into the database to conduct analyses of customers identified by defining areas on the map. With tens of millions of records in the database and a sophisticated data model, users needed a tool that could handle very complex queries.

"Many of our reports require numerous SQL passes. Most other tools simply could not handle the complexity without customizing the data model for each report," explains Snow.

The implementation was completed within six months, and now provides a powerful marketing tool for Cox. "We can identify candidates for new services based on their type of business, and generate a probable dollar value of their purchase decisions," Snow says, "and then we can see how they cluster geographically."

Third-party data is used to model data about likely purchase decisions. For existing customers, codes describing their type of business and existing services can help point to additional services customers may need. "For example, we might see an otherwise easy-to-miss opportunity such as a cluster of medical offices needing significant data services to transfer X-ray images" says Snow.

"We can even do analyses on buildings that contain multiple businesses with a wide variety of sizes and industries, such as seeing how much telecommunications spending goes on there," he adds. "That information is then run through a model that calculates the return on investment for building our network out to that location based on the realizable or expected value of the opportunity, the costs to extend the network and a risk-based cost of capital."

Those models can also help Cox avoid sending unwanted marketing materials to those who are not likely to be interested or are not economically serviceable. "It bridges the gap between strategic planning, marketing and sales execution," says Snow.

The system was designed to be accessible to both casual and power users. "Nearly 100 people in the field and at headquarters use the system," he says. "The casual user can click on a map, print it, render it as data and hand it to a salesperson. It's actually fun for people to use."

Power users, on the other hand, can go into MicroStrategy's Report Builder, create reports from scratch and do other sophisticated tasks. By making the system flexible, Cox was able to offer business intelligence to a wider variety of users without sacrificing functionality.

Partly as a result of the process of integrating MapInfo with MicroStrategy's BI tool for this application, the two companies have developed a formal alliance. "Many MapInfo customers layer the software directly over a database, with no BI component," says Tom Villani, VP of Global Alliances at MicroStrategy. "Now that the two products can work together seamlessly, these customers can conduct valuable analyses of their existing data."

For example, a company could look at any type of affinity buying pattern, such as the purchase of two major appliances, and then find out where those customers live as well as other demographic information, display it in map form and market a likely companion purchase. Or, they can begin with a target geographical area and work from the map back into the database, a powerful feature that Villani believes other solutions have not effectively incorporated.

Making sense of clickstream data

Almost at the other end of the spectrum from map data is clickstream data, which reflects activity on a Web site by recording the mouse clicks of all the visitors. Organizations are conducting more and more business on the Internet, and are therefore increasingly interested in deriving business value from that stream of data. But when thousands of users visit hundreds of pages, a huge storehouse of clickstream data is generated. Business intelligence solutions provide the means to convert that data into actionable information.

"Clickstream data warehouses can provide a wealth of information," says Mark Sweiger, president of Clickstream Consulting. "The most basic level of clickstream analysis is reporting traffic to pages on the site. For example, you can find out which pages are viewed most often by day, week, month or year."

Site owners are often surprised to find that traffic analysis identifies site pages that have never been viewed by anyone. If no one is visiting a particular page, the information it contains may not valuable, and the page can be dropped. Or navigation may need to be improved so the user can get there more easily.

"The next level of sophistication in clickstream analysis is user path analysis," says Sweiger, "which follows a user from one click to the next. Quite often, path analysis uncovers unexpected Web site navigation models." For example, many people may be entering a site from a bookmarked technical support page rather than from the home page. In that case, customers might not see critical information such as a new product announcement if it was placed on the home page. By using path analysis, site owners can make fact-based decisions on how to evolve their information and business models.

The third level of clickstream analysis is session analysis, which describes what users did during their entire visit to a site. Clickstream analysis of a search session, for example, can specify what terms the user entered and whether the search was successful.

"If users are spending a long time viewing a particular page," Sweiger says, "they have probably found what they were seeking, but if they immediately do a new search, they most likely have not." In the latter case, the solution may be a new information taxonomy, or use of a more consistent, standardized vocabulary in Web site content. Clickstream analysis can reveal which strategies will be best to solve the problem.

Clickstream data becomes even more valuable when combined with other corporate information. In one case, for example, Sweiger discovered that navigation from product literature to technical documentation was difficult. He asked the company's customer service representatives to classify inquiries received at the call center, and discovered that 20% of all calls were about how to find technical documentation.

"The reps knew about the problem but were not in a position to solve it," Sweiger observes. "BI provides the business justification for investing resources in a solution."

Cinergy is a diversified energy company that wanted to migrate as much customer service as possible to Web self-service. In order to help with the transition, Cinergy wanted to find out where customers were having trouble completing their tasks. The company had other Web analytics tools, but they were difficult to use, expensive to maintain or not capable of clustering multiple weblogs to integrate clickstream data from multiple servers. After checking out the alternative Web analytics solution options, Cinergy settled on NetTracker from Sane Solutions.

"One of our goals was to find out where in the sign-on process our customers were dropping out, either when they first registered or when they logged on for services," says Bill Davis, research manager at Cinergy. NetTracker helped Cinergy spot those dropout points. The company then was able to simplify the sign-on process; rather than having to go through multiple steps to get to the access point, customers can now sign on from the home page.

"The redesign was effective in reducing dropout rates," Davis notes, "and that helped us with the migration to self-service, which is much more cost-effective than call center customer service."

Although the most common reason for visiting the site is to get billing information, customers explore other content once they are on the site. Cinergy wanted to identify the content of greatest interest. The site provides information on a home weatherization program, how to reduce electricity use and how to find alternative suppliers. Popular information, revealed by NetTracker reports, is moved to a more prominent location on the site, and low-value content can be omitted in order to reduce maintenance costs for the site.

NetTracker's most recent product enhancements include the ability to use "event tags" to track "on-page" user behavior. That feature is useful in identifying such actions as how far down the page a visitor scrolled. NetTracker users can rely on the built-in analytical interface or use any of the leading BI tools as a front end.

"Organizations are starting to understand that they can do more than just measuring the traffic," says C. Decker Marquis, director of marketing for Sane Solutions. "The optimal use of clickstream data will come from combining it with other customer data from traditional data warehouses, so the ability to load all this data into a BI tool will be essential for some organizations."

OLAP plus text analysis = insights

Online analytical processing (OLAP) can present sales figures, detect trends over time and show many other quantitative performance measures, but it does not necessarily reveal why those patterns exist. Often, the reasons for the results of the analyses are embedded in unstructured data such as in customer e-mails, phone messages and comments in CRM systems. Or, they may lie within engineering reports about products. Unstructured data accounts for about 80% of corporate information, but it can be difficult to evaluate and interpret. Text analysis can reveal relationships in unstructured data and help explain the quantitative analyses.

Newly released Extractor 1.0 from Verity identifies patterns and concepts, aiding in automated classification of unstructured information. Once a structure is added, the information can be analyzed in a database or other enterprise application. Text Analytics 6.0 from ClearForest cuts across multiple repositories to detect relationships among various entities, including people, facts and events. It provides a solution for unifying structured and unstructured content. The analyses can produce positive business outcomes such as reducing spending on warranty repairs or increasing the speed of product development.


Judith Lamont is a research analyst with Zentek Corp., e-mail jlamont@sprintmail.com.


Search KMWorld

Connect

Buyers' Guide
Learn More in the Buyers' Guide!