The big bang of business intelligence

A recent New Yorker cartoon shows a doctor talking to his bed-ridden patient, and the caption reads, "It's a medical miracle that you made it through the last medical miracle." That cartoon pretty much sums up how telephone companies have weathered the storms of decision support analysis-changing technologies, dated skill sets and extensive organizational changes-and emerged triumphant in their ability to leverage business intelligence.

Part 1 in a two-part series Indeed, the competitive pressures are everywhere. Governmental deregulation and loosening restrictions are opening up markets and inciting an almost Darwinian struggle to stay competitive, let alone profitable. At the center of these battles lies corporate data, which has evolved into nothing less than a strategic leg up for telecom companies worldwide.

But telcos have learned the hard way that they need to walk before they can run. When it comes to deploying critical data to business people in the telco world, there have been almost as many false starts as successes. The key is incremental implementation and the payoff is sometimes miraculous.

Buzzword for the millennium? The advent of data warehousing technology across industries has revolutionized the way companies are selling their products, resulting in improvements as simple as obliterated query backlogs and as complex as predicting customer purchases. The term "decision support," or "decision support system," has historically been used as an analysis catch-all, reflecting the spectrum of data inquiry capabilities from simple queries to complex statistical models. The answers to even the simplest decision support system queries can result in far-reaching decisions about how, when and to whom to sell a company's products.

But decision support has evolved from straight question-and-answer to a host of different analysis techniques. As the number of data analysis techniques continues to grow and become more disparate, decision support has become only one of several ways that companies can use information to better run their businesses. This gamut of analysis techniques falls under the rubric of business intelligence.

Data warehousing lies at the center of this information whirlwind. Each of the regional Bell operating companies and long-distance companies has at least one data warehouse; many have several. In the case of an enterprisewide data warehouse, telecom employees from product managers to call-center staff can access the data warehouse to obtain important information about their customers, products, bills and networks.

Data marts, or functionally oriented data warehouses, target analysis to a particular business function. Telco business units are lining up for a piece of the data mart action by implementing their own data marts, separate but supplementary to their enterprisewide kin.

Figure 1 shows a simplified corporate data warehouse environment. Notice that there are several different data sources and several different types of users. These distinguish the enterprise warehouse from the departmental data mart used by a single functional area.

We've all heard the comparison between data, information and knowledge. Transforming the former two into the latter is what business intelligence is all about. For a company to truly merit the term "business intelligence," it must act on the product of its analysis efforts. Like the patient in the cartoon, taking sound business action based on meaningful information often involves one miracle after another.

The old standby As the director of information services for Infonet Corp., a global network services provider, Clark Murray has a lot on his mind. The company's adoption of data warehousing technologies is well underway, and both internal users and external customers are champing at the bit for data.

Murray's team has already identified basic business drivers for its data warehouse. What will they do once they've socialized decision support across the company?

"There are four primary areas that we feel will provide the greatest return on our data warehouse investment," Murray explains, "network performance, product profitability analysis, identifying sales opportunities and billing analysis." But the trigger was the need for timely and easily accessible performance reports, he says. "We find we still have numerous requests for performance reporting. We believe a data warehouse will help us provide better responses to these requests."

Murray's priorities are telling. Forget the hullabaloo over statistical extrapolations and what-if analysis-businesses need data to support day-to-day tactics. Infonet's customer focus has ordained the company's move into decision support by way of standard reporting, the bread-and-butter of data analysis.

Standard decision support has offered telcos speed and flexibility in areas such as revenue analysis, trouble tracking and strategic planning. Decision support queries allow business people-who might have previously waited for days or even weeks for answers-to access data right from their desktops. One telco required financial analysts to wait for a tape mount to get revenue information that was more than a year old. Depending on the backlog, this would take weeks. Now these same analysts can access that information in seconds.

Most software tools used to submit standard decision support enable users to "can," or package, reports, particularly those that are run regularly. This means the user submits a query with a click or two of the mouse.

However, while simple enough to ask, standard decision support queries can have significant influence on the business. For example, "List all business customers whose monthly calls have decreased by 20% or more."

The answer to this question can trigger many different business decisions, from offering discounts on products and services to augmenting network capacity in a given area. Certain actions can go so far as to improve customer satisfaction, or even prevent a customer from leaving. Regardless of the potential benefits, telcos could only recently gather this level of detail quickly and accurately.

One telco runs the following canned query weekly: "Display all cellular customers whose inbound calls were incomplete more than 10% of the time last month."

The answer to this question shows high-use customers who might need additional equipment. It can help indicate probable market competitors, suggest candidates for upgraded products, may validate suspicions about fraud, and may point out potentially dissatisfied customers. In short, it could earn the company additional business. As soon as business users find the information they are looking for, they might analyze the results with on-line analytical processing (OLAP) tools.

Standard decision support queries are underestimated. Telcos intent on keeping up with competitors, often try to run before they walk by deploying overly advanced analytics. However seasoned a telco's analysis staff is, business people-the end user majority-always return to standard decision support. Indeed, standard decision support targets the people running the business-not technologists-rendering it the most pervasive analysis type in a telecom company, and thus arguably the most valuable.

OLAP and multidimensional analysis Once simple decision support is available, new types of analysis can be conquered through OLAP. OLAP tools provide different perspectives on common data by permitting different levels of access. Furthermore, they allow users to organize their reports based on the way they need information to make decisions.

Consider the following standard decision support query: "Show me quarterly booked revenue for large business customers in the northern, northwestern and southwestern regions for 1997 and 1998." The answer to this query would appear in tabular output similar to Figure 2. It turns out that the revenue in the Northwest looks lower than expected.

The user may choose to examine another level of detail: "Show me the same data by district within the Northwest region." Because the Northwest region's District B has the lowest revenues, the user would want to examine the area's geographic breakdown to identify the root problem: "Show me the same data for District B in the Northwest."

Clearly the Cleveland revenues are lagging behind the rest of the district. By drilling down to more detail, the user can quickly track a revenue problem to a specific location rather than assuming the problem was company- or territory-wide.

OLAP tools are unique in that they make it easy for users to ask for the same data in different ways, in effect "slicing and dicing" it in order to mold the results.

A list of products and their sales dates doesn't do a life cycle manager much good; she more than likely wants that information by geographic region, area code or demographic segment.

Communications companies are using OLAP tools to make other consequential business distinctions. For example:

* Customer and product analysis. Identifying year-over-year (last year vs. the year before) or month-over-month revenues for current and prior time periods to track gradual changes in product sales or customer behavior.

* Industry behavior. Comparing different industries in terms of usage rates, product sales and revenues in order to customize and target-market certain products and services.

* Network traffic breakdowns. Looking at time bands of network traffic over specific routes to determine possible network upgrades or service changes. A common question is: If we add a new calling feature to a certain geography, how many switches will require upgrading?

* Call detail analysis. Inspecting use-including number of calls, minutes of use, time of day and origin and destination geography-in order to develop an understanding of call (and customer) behavior.

* Customer and product churn. Reviewing service cancellation by product and by customer to indicate whether a customer has chosen a competitive provider. Telcos may ask, "Have we lost any customers where our competitor has installed a fiber loop?"

As with standard decision support, end users can usually hit the ground running with OLAP. The tools are widely available and relatively simple to implement. Their value is that they are no longer reserved for blue-sky strategizing, but rather for laypeople dealing with day-to-day business issues.

More advanced types of business intelligence techniques don't necessarily share these characteristics.

Part 2, which will appear in the Feb. 8 issue of Telephony, will examine how one telco looked at old data in new ways using knowledge discovery.