Although they don't get around quite like Dr. Who, Travelocity.com's Paul Briggs and Caroline Smith manage a data warehouse that helps customers go virtually anywhere in the world.

 At a glance:

ï $193 million in sales for 2000
 ï #1 travel site after merger with Preview Travel
 ï Tickets issued in more than 90 countries
 ï Manages its data on a Teradata database using a two-node, 1.5 terabyte NCR WorldMark 4800 server 

Related Story: Data warehousing is going places

Just The Ticket

Travelocity.com books Teradata for virtually instant answers

by C.C. Williams

Travelocity.com, the Web's top travel site, promises that customers can "go virtually anywhere" in the world at the best fares, with just a few clicks of a mouse. Travelocity.com is reaching its own destination-a top-notch personalized online travel service-thanks to a "passport" from Teradata.

"Passport" is the name of the data warehouse system Travelocity.com is using to better mine the rich and growing vein of customer data flowing into the company daily through its Web sites, which offer customers deals on airline tickets, car and hotel rentals, vacations, cruises and more.

In just five years, the Fort Worth, Texas, company has grown into the world's largest online travel service in terms of travel bookings and monthly visitors. It boasts more than 27 million members, selling more than 13 million airline tickets and generating gross sales of $2.5 billion last year. But to stay atop the competitive online travel industry, Travelocity.com knew it had to find a way to better understand and serve its growing legion of customers, and more quickly answer increasingly complex marketing questions. So it turned to the Teradata Warehouse.

Paul Briggs, director of CRM, came to Travelocity.com with a background in marketing. Caroline Smith, senior manager of data warehousing, brought her IT experience to the mix. Together, they co-pilot Travelocity.com's data warehousing initiative, and they say it's been a smooth ride with Teradata.

Q: Where is the online travel business now? Is it growing and healthy?

Paul: It's growing very, very fast. Travelocity.com did about $2.5 billion in gross sales last year. We are the No. 1 site in the travel category, and the travel category is the largest e-commerce category. It's a perfect category for the Web because it's a virtual product. Even paper tickets are kind of going away. The tools we put in place really help to make the complexity of finding a good price and best price a lot simpler than the "old world."

Q: How did you choose Teradata to help with this effort?

Caroline: Back in 1999 - we had a data warehouse that was adequate for our size, but we knew that it wouldn't scale to meet our needs after we merged with Preview Travel (in March 2000). We knew that we were going to be a much larger company. We were going to have a lot more customers, and we knew that we were bringing in customers from AOL and other new avenues. And, with the system we had, we would not be able to scale like we wanted to; we would not have the speed that we needed.

Q: How much speed were you looking for, and what were some of the specific problems or issues with the older system?

Caroline: We weren't looking for a specific speed or capacity, just in case we grew faster than our estimates. We just needed the system to handle at least a terabyte of data, to be able to scale easily, to grow to any size that we demanded of it and to process faster than our current system.

Some of the specific issues with our older system were: daily loads were excessively long, (and) overhead cost was high. As far as overhead costs, we had to constantly manage file space in the old database, whereas with Teradata you just create a database, specifying the space needed for the entire database. You do not have to add or reallocate space every time you create a new table. Also, (the old system) did not scale, we had major outages for upgrades and tuning, complex queries took too much time and some queries (such as bail-out analysis) were impossible or would take days to run.

You have to have a database that answers the question that the internal customer doesn't know he's actually going to be asking that day. The customer will ask one question, and, after he gets back his response, he'll probably ask another question. It's kind of an evolutionary-type of process, and the questions build upon each other. You have to be ready to answer any question and be very nimble as a data warehouse. That's where a data warehouse differs from a regular transactional database. We needed some sort of technology that could scale properly and be fast enough to complete the daily data loads before the start of the business day.

Q: Travelocity.com probably saw a lot of pitches from a lot of companies. Why Teradata? What were the primary challenges that Teradata's data warehouse solution was intended to solve?

Caroline: We needed to act fast. We wanted to get a top three list of data warehouse solution vendors, and we went out to several of the research services, like the Gartner Group. After getting their advice, we came down to basically three players we were going to look at. One of them was NCR and Teradata. Coming from Sabre (Holdings, creator and part owner of Travelocity.com), we were already pretty familiar with Teradata because Sabre had been using Teradata for years. In fact, I used Teradata 10 years ago when I was a Sabre programmer.

But Paul and I were most impressed with (the fact that) unlike some of the other companies we talked with, right when we contacted Teradata, they were there; they were very excited about showing us what they could do. Within a month after contacting them, they actually had us out (to their facility) in San Diego with a copy of our database, and we were actually benchmarking our data on Teradata. This was even before the competition was able to do a debriefing for us.

Q: So it was a relatively quick process to decide to go with Teradata?

Caroline: Yes, it was over a period of a few months.

Paul: For a company their size, they moved more quickly than their competitors did.

Q: Did the conversion to Teradata go smoothly? What were the main challenges?

Caroline: The conversion went very smoothly. We really faced no major problems with it. Most of our problems were due to our juggling so many other projects at the same time the conversion was taking place.

Paul: It went pretty quickly. We had two or three people working for six months, and that was about it.

Caroline: We also underestimated the amount of time it would take to cleanse the data as it was being transferred over to Teradata. For instance, our old database allowed duplicate records in some tables, whereas Teradata did not. From conversations that I have had with other companies that have done a similar conversion, this underestimation of time needed for data cleansing is a common problem.

Q: How many divisions were involved in the conversion, and were there a lot of databases involved?

Caroline: We just had about three programmers working on it, and we had some assistance from Paul's group on the business side. It was done from our headquarters, and we didn't have any divisions involved. 

Paul: We were still a relatively small company, and when we originally started and launched a Web site we had to focus on having one data warehouse. When we decided to go with Teradata, we reconfirmed that decision to have one warehouse. All our data goes into it. We don't have a bunch of data marts; we don't have marketing data marts. It's all one warehouse so we can come at the data however we need to.

Q: How does Teradata's data warehouse solution fit Travelocity.com's overall business strategy? 

Paul: I'm not sure it's so much business; it's a question of business strategy. But we very much believe that you have to have one single view of the customer and you ought to have one central data warehouse. Our core strength, besides our Web site, is that we have 27 million members and all of that information needs to be in one place so we can best meet those needs. When we do market research with our members, they frequently say, "Why don't you personalize things more for me? You know where I live; you know when I come to your site. Why don't you use that information to help me out?" Our first response, frequently, is that we do. And we do try to send you information more relevant to you based on information you provide to us.

Q: What are a few of the major needs of your customers, and what insights have you gathered about them?

Paul: The reason we were four years old before we moved to Teradata is that, at the most basic level, the customer first needs a fast and easy-to-use Web site. I think we did quite a good job of giving the customer a lot of tools to be able to find the best fares and best travel for their needs. The other question is, "How do you take that to the next level by doing things that make the overall customer experience-whether it's e-mail or call center or on the Web site-more personalized and make things more timely and relevant?" That's where the warehouse really comes in, because we are able to pull things quickly. When there's a fare sale, we can turn around that communication to customers very, very quickly. It's all about having it in one space. It's not about only one insight we've gathered about our customers. It's about developing a much more thorough picture.

If you can run a query in a tenth of the time that you can on another system given the same amount of data, you can have 10 insights in the space where you would have had only gotten one. And that's what Teradata has allowed us to do. So instead of having a complex query running for half a day, it might run for a few minutes or half an hour. We'll have the answer we need, and then we can go back and refine that query - maybe we can get the answer in an hour instead of half a day.

Q: Last year, NCR said the Teradata system would enable Travelocity.com to get answers to questions previously considered impossible. Is that true, and what would be a typical query you'd ask of the system?

Caroline: With the Passport data warehouse, Travelocity.com will be able to provide services to our customers and gain insights about our customers' needs that we have never had before. A very good example of that would be next-page analysis. That would be when you look at all the pages that people are looking at for a particular time period. You want to see what people are doing on a particular page. Are they leaving the site, which is called "bailing out"? Do they continue down the (booking) path? Are they getting an error? We can get these answers quickly with Teradata. That's something we never would've dreamed of being able to do before.

Paul: That's one of the great things about Teradata. We have vendors frequently calling us offering to sell us tools to do bail-out analysis and all sorts of different services that generally revolve around pulling data out of the warehouse and putting it into data marts and then running analysis on that subset of data. Well, most of our data is in one place; it's all accessible, and it runs very quickly.

Caroline: Another example might be robot detection. There are a lot of people out there-hackers-who repetitively hit a site like Travelocity.com with "robots" to get information, such as airfares. We can detect that, trace it back to (the source) and stop the robots from using our site.

Q: How much did you consider privacy-protecting customer data-when doing the conversion?

Caroline: We take privacy very seriously here at Travelocity.com, and that's why we take an extra precaution as we build this data warehouse to keep it very separate from the transactional customer database and from the 
Web site servers.

Q: What's been the impact of the conversion on Travelocity.com's bottom line? 

Paul: It's difficult to separate what the data warehouse did compared with the effect of a better e-mail or sending it in a HTML format or improving the Web site. What we do know is that we look at the total cost of ownership of Teradata: the hardware and the software and the hours that are required to maintain the product and do the loads. We know that it isn't any more expensive (than the old system). In fact, in some cases, it may be less expensive than the system that we were on. Given the vastly superior performance we've gotten from Teradata, we know that it's been a good investment.

Q: Have you measured how much faster you have been able to serve customers?

Paul: We know that the same queries run a lot faster and the daily loads happen substantially faster. Daily loads, based on the same amount of data, decreased over 70%. One monthly report went from 11 hours on the old system to 20 minutes on Teradata.

Q: What additional CRM or data warehousing strategies are you planning to pursue?

Paul: We are looking at a lot of things. We continue to move toward that classic single view of the customer. We are probably pretty close. I'm not sure there is any company that really is there. But we have been on this journey for several years, even before Teradata. We're moving into different things in campaign management, data mining, predictive modeling. A lot of these things we might not have been able to do if we had remained on our old system.

Q: Database marketing combines IT and marketing. How do you see the developing relationship between those fields?

Paul: (Caroline and I) have been very aligned from the start. This whole exercise of migrating to Teradata was a joint project between marketing and product development , which is essentially our IT department.

Caroline: When we were looking at Teradata, Paul and I talked to other companies that have data warehouses, and the ones that are really struggling are those companies that don't have an aligned user group and IT group.

Q: Is there any advice you can give to another company considering a CRM/data warehousing initiative?

Paul: I think there are some basic things. You must focus on data quality; you have to have people that are dedicated to that, (to make sure that) changes to your Web sites are loaded correctly in the warehouse. That's what we have spent quite a bit of time grappling with. Our data quality is quite good. The other area is, you need to think about an agreement on service levels between different groups. What that means, for example, is the data warehouse will be available by 9 a.m. every morning with the previous day's data, 95% of the time.

Caroline: The best advice I can give to other people is to talk to other customers, find companies your size that have data warehouses and (ask about) their experiences. Learn by other people's successes and mistakes.

Q: How did the two of you arrive at Travelocity.com?

Paul: I was at Sabre for a year, prior to coming to Travelocity.com three years ago. I started in February of 1998. I was doing market research for Sabre and worked in business development here, then moved back into marketing. I then managed what we call the customer marketing group, which really focuses on managing the customer life cycle on Travelocity.com and getting the right message to the right customer at the right time. That's classic CRM.

Caroline: I, like Paul, was a Sabre employee. I joined Travelocity.com in December of 1995 when we were in the middle of development of Travelocity.com. We launched in March of 1996, so I was on the core development team. When I came over, one of my first jobs was to lead the technical marketing group. I (subsequently) became the manager of data warehouse group and just recently started up a new group called business intelligence.

Q: What do you find most challenging?

Paul: Most challenging is just trying to keep up with the pace of everything, trying to make the right decisions. We have just thousands of things happening every day, and we have a lot of e-mail campaigns going out the door and lots of other things happening. It's really about execution more than anything else.

Caroline: During the conversion of the data warehouse-while we were moving data from our old warehouse to Teradata-we were also working on about seven projects to add new sources of data to the warehouse. This was a real juggling act since we had to be careful to make the changes to the old and new databases until the cutover took place.

At Travelocity.com, we are constantly implementing new features and functionality into our Web site to provide more value to our customers. This impacts the data warehouse because new features on the site equal new sources of data for the data warehouse. Also, our traffic keeps going up, so we have greater volumes of data. Day to day, we have to make decisions about evaluating and prioritizing these new sources of data. This can get quite hectic since everything is always a No. 1 priority around here. T

C.C. Williams is a New York-based writer who covers business, finance and technology issues.

Teradata Magazine - Q2 2001