Onsite: Alaska Airlines Takes Direct Route to Analyze Customer Data

Plenty of companies have difficulty connecting information silos, and it's no different in the airline industry. At Alaska Airlines, the ticketing and reservations applications are built on one system. The Mileage Plan frequent flier program is built on another. And the customer relations department has yet another system to help it address extreme delays, lost bags and complaints. The challenge was pulling together data from across these points for a holistic view of the customer.

The airline wanted to increase loyalty and boost revenues with personalized promotions and Web site interactions, but without complete knowledge of a customer's history and future activity, it was impossible to target campaigns based on customer-relevant insight.

"If we know a customer has a higher propensity toward vacationing, and it doesn't dilute our revenue, we might offer better discounts on vacation packages, such as ski trips," explains James Archuleta, director, customer relationship management (see "Driving Force," below).

The company's first thought was to use integration tools, such as enterprise application integration, to join transactional systems. But a study revealed that would be a difficult path because the airline's systems were so disparate, with proprietary systems on different platforms, including legacy IMS and COBOL as well as the airline industry's hosted Sabre ticketing system.

An ETL (extract, transform and load) process could tie the data together in star schemas through conforming dimensions, but Archuleta says the company lacked the business intelligence and reporting infrastructure required to support that approach. Existing BI (business intelligence) technology included Brio tools and home-grown SQL systems, but without a unifying metadata layer, these tools could only analyze one silo at a time.

In early 2005, the airline finally settled on building an OLAP platform.

"We wanted something robust that would let us keep our legacy systems in place--a nice data warehouse or federated data mart, as we like to call it," Archuleta says. "For that, we needed tools that could handle those structures."

Alaska Airlines chose Siebel Analytics. The airline wasn't using Siebel CRM software, but like many customers, it was attracted to the analytics product for its strengths in providing federated access to disparate data sources. (Indeed, Oracle considered Siebel Analytics one of the plums in its acquisition of the CRM vendor, and the software is now at the core of its recently launched Oracle Business Intelligence Suite.)

It wasn't a hard choice for Archuleta, either, having helped launch the product at Siebel before joining the airline in 2004. Still a champion of it, Archuleta says Siebel Analytics offered "the strongest interface for our business users, the strongest dashboards and the strongest functionality around metadata and implementing business rules." It wasn't the lowest-cost choice up front, he admits, but he adds that low training and deployment costs made it best choice for total cost of ownership.

Implementation took about six weeks and involved modeling the data from disparate sources and databases using the star schemas. It was critical to be able to pull data from both legacy systems and the hosted Sabre ticketing information. This included time-variant, year-over-year and month-over-month data that would be essential to understanding trends.

Next, Siebel's metadata layer was used to implement business rules, aggregation rules and drill rules so business users wouldn't have to think through those steps. "When someone wants to drill from year to quarter to month to day, those drill downs are already defined through the metadata layer, and it really allows our executives to get to actionable strategy on the fly," Archuleta says.

Executives now look at aggregates by segmentation of the loyalty program customers, and as they drill down they can spot customers who stopped flying with the airline based on ratios before and after a given date. They can then investigate whether there's a corresponding history of flight delays, lost bags or complaints. "We actually alert our executives with these lists, and they'll make calls to high-value customers to try to retain them," says the CRM exec.

In another example, the airline can examine whether certain flight segments in the route structure are causing customer attrition. "We can anticipate many things based on history, such as needing more baggage handlers over the holidays," Archuleta says, "but we're now identifying other cycles as well, such as pre-spring break and post-spring break demand and variations in ticket demand by city." The airline now sets staffing levels more intelligently based on analysis of the latest ticketing trends.

The biggest payoff, however, has been in helping Alaska Airlines deepen customer relationships and improve customer loyalty with everything from e-mail newsletters and affinity credit cards to vacation packages and Mileage Plan membership. "Previously, our campaign weren't geared toward relevancy," Archuleta says. "But tying everything together has helped us be more focused, more efficient and more cost-effective in our promotions."

As director of CRM at Alaska Airlines, James Archuleta sets the road map for all customer-facing initiatives from both a technical and a business process perspective. When he joined the airline in 2004, he set his sites on creating "process-driven analytics," but knew that meant joining disparate systems at the metadata level.

"A lot of companies have metadata but don't realize its potential," Archuleta says. "Taking a metadata-driven approach helps you understand the relationships between disparate data sources and standardize your business rules. With any product, it's critical to understanding business rules and how your data relates to these rules."

It's too easy to attribute Archuleta's choice of Siebel Analytics as loyalty to a product he helped introduce during his tenure at Siebel. But Archuleta's focus on metadata predates those days. Previously, Archuleta spent 10 years at Motorola Semi-Conductor, culminating in the position of senior staff scientist directing the integration of disparate data sources via a metadata layer.

To form a clearer picture of the customer and key operational issues, Archuleta recommends "taking a step back, looking at your entire warehouse and understanding the data relationships on a logical level." That's not only the key to understanding data integrity issues across the organization, it's the key to clear communication and consensus between IT and the business community.