Cathay Pacific Airways is a $3.6 billion airline with 14,500 employees serving 10 million customers a year in 50 countries on five continents. Recently, we decided to invest in a customer information system (CIS). The new system was intended to unify our separate databases into a single data warehouse and provide information enabling us to match corporate strategies and products more closely to the marketplace.
Our project was triggered by a need to understand more about our customers. The challenge for airlines, as it is for many other businesses, is to access and integrate the customer information scattered around many different legacy systems. Our goal was to use a data warehouse as a decision support system to make better decisions around customer and revenue management. Since the resources required to build a data warehouse are substantial, we began with a pilot project in early 1997. When that proved successful, we built a production system beginning in February 1998 and have upgraded it several times in the past year.
The 1997 pilot project consisted of a reduced-scale data warehouse containing only a few pieces of information with specific business application and a measurable business benefit. Running Sybase 11 on an IBM RS/6000-J40, the pilot was intended to demonstrate to corporate executives that the data warehousing concept actually worked. While the concept proved successful and funds were allocated to build a production-scale data warehouse, execution did not measure up.
In the pilot project, with Sybase and IBM, we had a problem with poor performance. We decided to go out and search for another platform. Beyond the key factor of raw performance, evaluation criteria included company profile, middleware, scalability, local support and price. One of our key criteria was the overall performance and scalability of the production system over a projected four-year period. After an initial process of elimination, we were left with two competing solutions: Oracle8 running on a Sun Enterprise 6000 server and IBM SP/DB2 running on an IBM RS/6000.
Benchmarking was done at Sun and IBM testing facilities, and both solutions met our minimum performance requirements. However, when we factored back in all our evaluation criteria, Sun won out as a total solution on the combination of benchmark and price/performance. In particular, the Sun StorEdge disk array solution was very competitive in price/performance against IBM's SSA storage. Sun's RAS features were also considered important in tilting the decision in Sun's favor. Based on the benchmark performance test and other issues not least of all price and support we made our decision to choose Oracle and Sun. That was the best combination that suited our business needs.
Our choice was confirmed when we saw the improvement in data loading performance. Previously, it took about eight to nine hours to complete all the data loading into our data warehouse. After switching to Oracle on Sun, it now takes only three to four hours.
We made our first purchase of Sun hardware for the data warehouse in February 1998, buying a Sun Enterprise 6000 server and Sun StorEdge A3000 disk array with approximately 350GB of storage capacity. The next 10 months would see two upgrades, plus an additional purchase of Sun equipment to create a data mart.
The new data warehouse doesn't replace any of our existing operational processing systems. We take feeds from these systems into the data warehouse, which has a data model that allows us to integrate that information. What makes the data warehouse an attractive business investment is that you don't have the risk and hassle of trying to build all sorts of links between your disparate legacy systems. It allows you to integrate the information without having to disrupt the operational requirements.
In May 1999, we made our first upgrade to the data warehouse, adding about 500GB of new storage capacity. We purchased three Sun StorEdge A3000 disk arrays with 378GB of storage in total. We also bought a 128GB Sun StorEdge A5000 disk array for use in downloading the marketing information data tapes (MIDT) data each month. The Sun StorEdge A5000 disk array serves as temporary storage when we load about 40GB of MIDT data each month for processing. We then aggregate that data and put it into the data warehouse. Fibre Channel technology in the Sun StorEdge A5000 disk array provides us with the high throughput we need to load the MIDT data rapidly, and its 128GB storage capacity accommodates both current and future needs. The Sun StorEdge A5000 disk array also fit neatly into the Sun Enterprise 6000 server cabinet in the data center, conserving space.
The business applications in our data warehouse fall primarily into three areas: revenue management, target marketing and customization of products and services. Since the production system went live in August 1998, clear benefits are already apparent in both increased revenue and reduced costs. When we went into this we were expecting greater gains from revenue generation, but we were surprised by the big cost-saving benefit.
The biggest advantage our data warehouse solution has given us is that we're working more efficiently, with higher productivity levels. Apart from knowing the customer better and making better decisions around what the customer wants, there are also internal opportunities for working smarter and lowering costs. I think that's a key learning point that was a little bit unexpected.
The next step for us is to move the customer information to our front-line staff to help them make better decisions on the spot. How do you get the knowledge out that this customer is a very high lifetime-value person with a particular affinity for a specific product? How do you get that information in an online mode to the right person at the right time? That's the next big challenge.
In making the corporate case for a data warehouse, we advocate a two-tiered approach. You need to be able to go for the more strategic, longer-term benefits, such as the customization of products and services and target marketing. The ROI is bigger, but it takes longer. So you also need to have some tactical, lower-hanging fruit, such as revenue or fraud protection. If you go into this just trying to get the long-term strategic benefit, you probably will lose executive support because they won't see the results along the way. And if you only go for the lower-hanging fruit, then you are turning your back on the really big gains that give you competitive advantage.
Register or login for access to this item and much more
All Information Management content is archived after seven days.
Community members receive:
- All recent and archived articles
- Conference offers and updates
- A full menu of enewsletter options
- Web seminars, white papers, ebooks
Already have an account? Log In
Don't have an account? Register for Free Unlimited Access