In its image

Many industries lack sufficient customer data to conduct meaningful analyses. Luckily, this is not a problem for telephone companies. On the contrary, the key challenge for carriers is to integrate and consolidate their massive amounts of information. Their data is often inconsistent, redundant and resides on multiple databases, mainframes and Unix servers.

Once the warehouse is designed and populated, telcos have several application choices. A recent survey of 28 carriers worldwide revealed that applications driving the use of data warehouses include target marketing and lead generation, product profitability analysis and promotional campaign management (Figure 1). Topping the list was customer profiling and segmentation.

Proper customer profiling lets a telco find, exploit and hold onto valuable customers. It can also uncover untapped market opportunities. For instance, a carrier might discover that Latino customers are one of its most profitable segments and develop a campaign to help market to this previously overlooked niche.

With a single view of customers across several markets, lines of business and services, carriers also can determine a customer's value down to the bottom line. For example, a customer may be spending $1000 a month on long-distance services but almost nothing in wireless. Should the wireless bill become delinquent, the warehouse alerts the carrier so that the wireless service isn't automatically disconnected.

Of course, the definition of a valuable customer will vary. Each business unit tends to color its interpretation according to its own point of view.

Data warehousing helps overcome this parochialism to arrive at a broader understanding, one with a more objective way of assessing customer value independent of viewpoint. A new service may be boosting revenues in the short term, but the return on investment may fall short.

Churn management Interexchange carriers, wireless carriers and cable operators are struggling to manage customer churn.

Cellular industry sources estimate that the cost to attract a new cellular subscriber today is about $400. The average cellular bill in the U.S. has dropped from $125 in 1990 to about $60 today, so a cellular operator won't turn a profit for seven months. Thus, the estimated 25% to 30% annual cellular churn rate can wreak havoc with the bottom line. Wireless carriers are forced to take drastic steps to prevent a customer from dropping service or defecting to a competitor. Here, too, the data warehouse is very useful.

Customer retention campaigns begin at the data warehouse. Using profile data, analysts model former customers to identify the patterns that led to their defection. Current customers are fitted to these models to identify likely future defectors; preventative measures can then be taken to keep them in place.

Poor handset transmission quality is a major cause of cellular churn. Thus, wireless carriers are mining their warehouses to find correlations between different brands of handsets and customer churn. They also are exploiting their warehouses to measure the profitability of customers attracted by different cellular distribution channels.

Not long ago, direct mail and target marketing were foreign terms. Telcos now view direct marketing as a gold mine of unexplored opportunities.

Warehouse systems are not just for business analysts. Programs are desperately needed to help put valuable warehouse intelligence in the hands of front-line customer service representatives. The warehouse can pinpoint valuable cross-selling opportunities and point to appropriate direct mailing promotions based on a predefined level of business or a particular customer profile.

A data warehouse can dramatically improve the quality of direct sales calling lists. The ratio of sales per salesperson, for example, can double-or even triple.

Data mining is a processing-intensive warehouse application that combs through reams of data seeking patterns and rules that can help predict customer spending and other behavior. Combining simple association rules with data mining techniques, a warehouse enables market analysts to discover rules linking dozens of services and customer groups.

One such uncovered rule might be that 80% of household subscribers who have both cellular and call waiting service also have voice mail. The opportunity to uncover other profitable relationships and otherwise undetected opportunities is endless.

Carriers also are beginning to use their warehouses to address network infrastructure problems.

Performance monitoring is the chief concern in network applications. Here, the warehouse collects data on thousands of network nodes and alarms to establish historical trends and pinpoint future investment in network infrastructure.

With decision support applications proliferating about the warehouse, data storage requirements of telecom warehouses will expand rapidly in the years ahead. Although only 26% of carriers in 1995 had warehouses of 500 Gbytes or more, that percentage is expected to increase to 63% this year.

One reason for the increase is that warehouses originally filled only with summarized data are now increasingly loaded with extraordinary volumes of detailed data. For example, a local exchange carrier that today maintains only one month's worth of detailed records on-line soon may store five months of detailed data in its warehouse-and keep highly summarized records of data for six months or more.

As decision support analysts obtain warehouse sophistication, some carriers plan to increase warehouse capacity simply because the number of queries would skyrocket as decision support analysts drill down-in many cases to the individual call record level.

Growing pains Telcos can choose from a range of data warehouse platform options, including mainframes, symmetric multiprocessing (SMP) systems and even single-processor computers. Telcos often prefer SMP architecture for its low cost and easy development environment.

System features that telcos consider important in their warehouse environment include highly scalable performance, database and computer platform interoperability, and best price/performance value (Table 1). Combining these systems with the leading relational database software packages such as Informix, Oracle and Sybase provides the database and platform interoperability telcos need.

More than one telecom warehouse project has faltered because planners failed to anticipate warehouse growth-and thus were stretched beyond the capacity of its servers. Service providers also are faced with this problem. For cost reasons they need to start small but know that their platforms can grow sufficiently to handle storage requirements in the terabyte range.

WorldCom was faced with such a dilemma. By the end of 1994, WorldCom-formerly LDDS-had suddenly grown into a telecom powerhouse. Having acquired WilTel and IDB Communications, WorldCom was one of only four U.S. carriers boasting a nationwide fiber optic network.

WorldCom knew that whatever warehouse platform it chose, it had to be able to keep up with its steep growth curve. Beginning in late 1994 and implemented within six months, the company's initial warehouse was built around two Sun SPARCcenter 2000 machines-one with 16 processors and one with 12, plus 2.5 Gbytes of RAM each. A battery of SPARCserver 10s and 20s also served as upstream data collectors off DSC central office switches.

Since then, WorldCom's warehouse has grown rapidly. Today, virtually all of WorldCom's call detail records are available for real-time queries within 10 minutes. Such rapid data availability is essential not only for fraud detection, but for data warehouse applications such as call tracing and switch network trouble detection and repair. As many as 100 PC/Windows computers can simultaneously use forms developed by WorldCom to perform ad hoc queries connected to the warehouse via a SPARCserver 20 Telnet local area network manager.

The textbook model of an all-knowing global warehouse manages informational data across an entire enterprise. Although that warehouse model suits some industries, it certainly does not fit telecom. Rather, carriers are deploying local data marts that often do not integrate with a single enterprise repository.

As a result, the short-term trend is toward multiple, decentralized warehouses and data marts.

Large carriers are inclined to buy warehouses for individual divisions, especially those with dynamic marketing organizations. With highly diverse operations and international ventures in progress, the philosophy seems to be to get value from a data warehouse today and worry about consolidation later.

But a centrally managed warehouse that crosses business units also can be extremely beneficial. For example, telcos are urgently targeting a growing customer segment-telecommuters and business professionals working out of the home. Carriers find they can no longer limit the definition of business customer to those who work at a business address and pay business phone rates. Telecommuters and other such hybrid customer groups are a strong argument for building a cross-business unit warehouse.

Local data marts also serve as useful vehicles for delivering enterprise warehouse data to end user departments. These marts are often owned and managed by the user group. Each mart houses only the information required by its own community of users. The information needs of small business marketing, for example, contrast sharply with those of consumer marketing.

Corporate culture has a lot to do with how warehouse data is deployed to end user groups. The information technology shop at some carriers maintains strong control over the warehouse architecture, while others adopt a more free-wheeling, entrepreneurial environment.

At MCI, for example, the key challenge was ensuring that the warehouse could provide diverse business units with control over how data was interpreted and manipulated. The traditional top-down approach to data warehousing was too broad to satisfy the unique business needs of various units, especially within such an all-encompassing data warehouse project. As a result, MCI's data warehouse architects decided on a delivery approach that focused on the specific requirements of each business unit.

In 1995, MCI's small business sales unit recognized that a modern data warehouse was critical to maintaining a competitive advantage. MCI therefore created its small business on-line lead database. At its center lies an eight-processor data warehouse server running Sybase, which is updated daily with millions of records from dozens of sources.

More than 100 external data sources and 35 internal MCI sources feed the database. Load volumes can range up to 2 million records per source. The data warehouse resides on 180 Gbyte storage arrays.

PC-based users employ a custom-built graphical user interface-based on the PowerBuilder application development tool from Sybase's PowerSoft subsidiary-to access records and generate leads via an intermediary SPARCstation 4.

To ensure around-the-clock availability, subset databases are spun off for read-only purposes while the main database is being updated or backed up.

The telecom industry is a prime example of the changing business environment driving the rapid development of data warehouses. The key is for service providers to embrace technologies that enable them to set up their warehouse today, but provide the flexibility and scalability to support their growing business and data storage needs.

Rod Tansimore is Manager of Data Warehousing and Decision Support for Sun Microsystems Inc., Mountain View, Calif.

Pay attention to data integrity. One of the biggest stumbling blocks when implementing a warehouse is ensuring that only clean and unduplicated information is contained in it. Data warehouses fail when management's attention is focused on the warehouse system instead of how the data will be used.

Think scalability. The experience of telephone companies that adopted warehouses early shows that the warehouse grows dramatically as a telco gains experience and its users increasingly require greater levels of summary and detailed data in their warehouse environment. A modular server design makes it easy and cost-effective to add disks, memory and even central processing units. This expandability permits flexible growth. As businesses change, customers can add, reconfigure or upgrade workstations, servers and other hardware-paying only for as much power as is needed today.

Stick to the emerging telecom warehouse standard, Unix. The move to open systems in telecom warehousing will be dramatic over the next several years. Although only one-third of warehouse computer processing was done on open systems in 1995, according to Technology Research Institute, 93% of telecom carriers favor Unix in future warehouse implementations. Unix scores high because of its reliable track record and its widespread support in telecom network systems worldwide.

Choose vendors with a track record in telecom. This could include suppliers that have worked with warehouse integrators, as well as database vendors and telecom business systems integrators.

Build on small successes. If analysts and other users in sales and marketing organizations buy into the warehouse program and quickly realize benefits, the program's success will sell itself and generate word-of-mouth momentum in other departments.

A good place to start is with a pilot project of limited scope with a duration of no more than three to four months. This pilot application can then be scaled up to full production mode and use off-the-shelf software packages.

Learn something about target marketing from other industries. Although target marketing is new to telecom service providers, direct marketing experts have delivered proven results for years to companies in the packaged goods industry. Carriers need to tap into the expertise that resides outside of the telecom industry.

Train users in the new on-line analytical processing tools. To overcome resistance and win the support of front-line analysts and users, carriers eventually will want to build sufficient on-line analytical processing training and support costs into their warehouse investment.