It's Show Time
A savvy reseller pushes all the right buttons for MovieFone
By Christy Walker
Sm@rt Reseller
H el-lo. If you know the name of the reseller with which you would like to work, press '1' now."
So MovieFone, which provides a national movie listing guide and ticketing service, pressed 1. It chose CSI Inc., and the systems integrator helped it secure faster and more comprehensive access to its customer data.
What MovieFone really was looking for was a more efficient method to capture and analyze customer statistics. Solving that problem became a job for CSI's Center for Data Warehousing.
Before CSI arrived, MovieFone had a dedicated PC to create a summary table that would then have Foxpro reports written against it. "If the dedicated PC went down, the load process would need to be restarted," says Glenn Peipert, CSI's chief operating officer.
Although it was collecting information, MovieFone was slow in getting its business managers the information they needed to push potential sales and market products appropriately.
"We had an X-based format in a Clipper system. It was like a home, self-developed report viewer," says Irwin Miller, Moviefone's VP of computer systems development. "We didn't want a DOS-based front end [any longer]. We couldn't do data parameter reports, we couldn't drill down."
A wholly owned subsidiary of America Online since May, MovieFone had a stockpile of cumbersome customer data that it wanted to turn into dollars. By the time CSI met the New York-based movie enterprise, the company had outgrown its system.
Known for its 777-FILM phone line and MovieLink Web site, MovieFone handles listings and ticket sales for 14,000 movie theaters in more than 35 major cities in the United States and Canada. As users press their keypad to navigate movie listings and other services provided by MovieFone, the company works behind the scenes to collect customer demographic information, preferences and purchasing patterns.
And a mountain of information it is -- MovieFone receives roughly 70 million calls a year.
Who's Buying Those Tickets?
CSI set out to help MovieFone design and build a data mart that would cull customer information, make it readily available to its marketing and sales departments, and enable managers to make strategic decisions about their business.
"We needed to provide information quickly to our management and enable ad hoc drill-down functionality," says MovieFone's Miller. "We needed to know, for example, how many people called Los Angeles [MovieFone] during a certain period, how many tickets were sold for Titanic and how it was broken down demographically."
New potential revenue was the primary goal. Precise information analyzed in a data mart could be sold to an exhibitor, for instance, looking to build an artsy theater in a city. MovieFone could sell the investor a breakdown of people by zip codes that request movies of that ilk and pinpoint the best location for the new theater.
"We could break down films by zip code, and our research department could then draw a map to highlight where [relevant viewers] were. We could do market research quite easily," says Miller. Similarly, market research and customer trends could be used in-house by MovieFone employees to direct incentive programs or telemarketing campaigns.
What Works
CSI first needed to determine the best data warehouse or data mart infrastructure for MovieFone. The integrator walked MovieFone through questionnaires and helped it narrow down options. As part of its services, the Hanover, N.J.-based reseller "test drives " data marts and data warehouses to ensure that the choice of tools is appropriate for each client.
MovieFone was interested in building a data mart using Sybase Adaptive Server IQ. The Sybase server enables corporations to perform data analysis and ad hoc queries. CSI needed to qualify what MovieFone wanted and determine if it truly was a viable candidate to use IQ.
"Sybase IQ is an excellent tool when companies are doing queries and trying to return counts, like how many customers listened to a particular preview," says Peipert. "But it's not phenomenal at pulling 20 particular records out of millions or retrieving information out of the database. It really depends on how the customer is going to use IQ."
CSI performed a test and loaded up a couple of million rows of data on Sybase IQ, then ran ad hoc queries against it. It passed the test with flying colors.
The next issue was finding an appropriate business intelligence tool that could query the data warehouse information as well as provide self-service access to end users. CSI and MovieFone looked at a handful of tools.
"[MovieFone] wanted to know the most appropriate business intelligence tool that would profile ease of use and interface," says Peipert.
As an independent reseller, CSI has a five-page matrix of questions it asks tool vendors. The answers are analyzed in a way that helps to determine--based on the different features and functions--which query tools are appropriate for an individual client.
"Sometimes the issues [for our customers] are financial stability of a vendor. If it's a start-up tool vendor, it may be a great product but sometimes it gets knocked out of the running," says Peipert. "We usually narrow it down to two or three tools, and at that point, we prototype the client's data."
MovieFone wanted a tool with a GUI that is easy to use and easy to maintain. BusinessObjects Inc.'s BusinessObject query tool fit the bill.
"We were looking for something that could easily build canned reports and would be offered to our users via a simple menu," says Miller. "We also didn't have a database administrator on staff yet, so we wanted a data mart that could easily expand as we pumped more data in there."
Beware of the Gotchas
Although the MovieFone project was completed in less than 120 days (for a fixed price of $75,000 plus products), in all data warehousing projects there are obstacles to be wary of and issues that arise. One hurdle that often needs to be surmounted is incompatibility between new versions of software and hardware, says Peipert.
"This is a classic fact of life," he says. "Our methodology is that when we build a data mart, we build an open solution. We don't endorse 'turnkey' solutions."
For example, a new version of IQ may not yet be certified with a specific business intelligence tool. Similarly, problems arise when an integrator installs a new version of a database, and the ODBC drives still have to be certified with it.
According to Peipert, a reseller must fundamentally plan for that incompatibility and schedule work to resolve such issues, on average one to two weeks. Another option, he adds, is to work firsthand with the vendor and make changes to the soft code. But if all else fails, work with the older version instead, "because vendors always come out with patches or a version that fixes all the problems," he says.
Another potential minefield is infrastructure design. The design of a data warehouse can make or break the system. Design issues range from aggregate data vs. raw data, to star schema representation vs. aggregation.
MovieFone had millions of rows of data. It was apparent to CSI that if a user wanted to do a simple query, it would be futile to go against all that data at once. It is always preferable to report against summary data, says Peipert.
In addition to aggregation of data, there are star schemas. The popular star schema--or representation of data in the form of a star with the fact in the middle and its dimensions surrounding the star--means performance benefits. However, certain databases support star schema, while others are optimized to support it. Determining which design or combination of designs are most beneficial to a client can be a very complicated process.
"Resolving this obstacle requires knowledge of the products [infrastructure] selected, volumes of data and reporting requirements," says Peipert.
The solution? A lot of planning needs to be done early on to figure out what types of queries users will ask the system.
"We work with our clients to identify current reporting requirements. We then work with them to discuss future requirements, which sometimes requires us to suggest appropriate views of their data," says Peipert. "We then work with them to identify hierarchies within the data ... this will also tell us how to properly architect the database."
Peipert adds that integrators should never start to architect the database until all the tools have been determined. "We worked with so many data warehouses that we've learned the hard way," he says.
Upcoming Previews
The future of data warehousing is blooming, according to Framingham, Mass.-based International Data Corp. IDC predicts total data warehousing market sales (including software, services, servers and storage) will grow at a 20.5 percent CAGR, surpassing $29 billion by 2002.
Because corporations increasingly are emphasizing the value of analyzing business trends as a complement to operations and as an enabler for strategic planning, investments in software, servers, systems and storage for data warehousing will continue to rise, says IDC analyst Jacqueline Sweeney. "There is increasing pressure on corporations to do more customer intimacy, and to do that, they need a data warehouse," she adds.
MovieFone is doing its part to add to this total data warehousing revenue. Besides the new software, MovieFone went for a full hardware upgrade and runs Adaptive Server IQ 11.5 on a dual-processor Dell Computer NT Server, as well as the Business Objects 4.1 query tool on six desktops. The plan is to grow to two dozen desktops and Web-enable the system.
Experience is the best teacher. With CSI's foresight and accumulated knowledge, the reseller was able to put in place an active data mart in less than 120 days. Success of the MovieFone data mart will ensure that moviegoers get what they want, and enjoy the show.
So what's playing at the theater this weekend? MovieFone's new data mart has the answer.
Christy Walker is a freelance writer, based in Cambridge, Mass.
Smart Steps
Obstacle 1: Incompatibility issues between new versions of applications or hardware
Tips
- When building a data mart, do not endorse "turnkey" data warehouse solutions.
- Fundamentally plan for incompatibility and leave time in schedule to resolve such issues, on average one to two weeks.
- Leverage partnerships with vendors. Development teams often make changes to code and will redistribute it to you.
- If the new version of software gives you too much grief, use the old one, instead.
Obstacle 2: Determining the best database design for the client
Tips
- Acquire knowledge of the products (infrastructure) selected, volumes of data and reporting requirements.
- Plan ahead. Always try to understand the business and its analysis needs (including future requirements) first.
- Identify hierarchies within the data.
- Assess if a simple aggregate, a star schema or a combination of both is most appropriate for the client.
Obstacle 3: Data transformation across global regions or from various applications and storage devices
Tips
- Have the right people involved, including business managers and IT administrators
- Plan for it to take longer. The length of the data transformation process all depends upon number of feeds and complexity.
- If the system(s) is outsourced, and those who wrote the app are not in-house, it will fail.
Article appeared online in ZDNet - September 6, 1999
