Selling custom ring tones and other downloadable content for mobile phone users requires staying in tune with the market. ABS-CBN Interactive (ABSi), a subsidiary of the largest integrated media and entertainment company in the Philippines searched its transactional data for hints on what to offer users in cross-selling value-added mobile services. But the searches took days and didn't provide customer-specific recommendations. ABSi wanted to increase its revenues from higher response rates on its marketing initiatives. Working with dB Wizards, a Microsoft Gold Certified Partner for Business Intelligence, ABSi deployed a data warehouse running on Microsoft SQL Server 2000 and a recommendation system running on the beta edition of Microsoft SQL Server 2005. The result is more accurate and personalized service recommendations to customers, and a doubling in response rates from marketing campaigns.
Situation
Mobile phone users in the Philippines love custom ring tones and the fun of receiving services including playing wireless games, reading celebrity journals, downloading picture messages, and participating in interactive television promotions and Short Message Service (SMS) chats. A leader in meeting the needs of this market is ABS-CBN Interactive (ABSi), a subsidiary of the largest integrated media and entertainment company in the Philippines.
With millions of unique users per month, ABSi is passionate about staying in step with its users and providing them with the most relevant experience possible. Toward that end, the company studies customer buying patterns to determine trends and predict the best future offerings.
Through agreements with wireless carriers, ABSi can send users a number of messages weekly offering new ring tones and other value-added services. The key objective of the company is to increase revenue per user by maximizing the rate of response and making optimal use of its allotted text messages by making the offerings to the users as relevant as possible.
In an effort to enhance sales of ring tones, the company studied online transaction processing (OLTP) data to predict the best ring tones to offer. OLTP data was analyzed by running Microsoft SQL Server 2000 (part of Microsoft Windows Server System integrated server software) SQL queries against it, and then copying the results into a Microsoft Office Excel spreadsheet for analysis. Management reports generated with this system took at least three days to create. This meant the reports were never up to date, and because of the manual processing involved, were prone to error. For customer recommendations, the system provided helpful information, but could take nearly a week to complete and resulted in general offerings that would be made to broad groups of customers.
"With our old system there was a lot of gut feel to it," says Grace Cunanan, Technical Specialist at ABSi. "We'd look at the data and choose the same list of five songs to send to everyone who had recently ordered a ring tone."
ABSi decided it needed a data mining solution to create customized suggestions that are relevant to each unique user, in hopes of enhancing the response rate beyond the existing level. The company needed a data mining solution that was both fast and accurate.
Solution
Working with dB Wizards, the first Microsoft Gold Certified Partner for Business Intelligence in the Philippines, ABSi has deployed a data warehouse running on Microsoft SQL Server 2000 and a data mining solution based on the beta edition of Microsoft SQL Server 2005. "Code named Project 88, the solution generates personalized recommendations for each user," says Poch Reyes, Technology Solutions Manager for dB Wizards. "We focused on the ring tone service users for the initial release of the recommendation system. So far, we are seeing remarkable results, and ABSi management is thrilled to implement this to other service groups as well."
ABSi plans to bring increasingly more information sources into a data warehouse to support data mining. The Phase 1 architecture includes:
Data Warehouse. The data warehouse uses SQL Server 2000 data transformation services (DTS) to import OLTP data from the Short Message Service (SMS) servers that handle ring tone and other transactions. Millions of transactions are imported each day, using 50 DTS packages. The data warehouse runs on a single instance of SQL Server 2000 on the Microsoft Windows 2000 Advanced Server operating system. The data warehouse is hosted on an Intel-based server with two 1.13 GHz processors, 4 gigabytes (GB) of RAM, and 1.2 terabytes of extended storage.
Reporting. SQL Server 2000 Analysis Services is co-hosted on the data warehouse. Analysis Services is used to create three multidimensional cubes and a number of virtual cubes for reporting. SQL Server 2000 Reporting Services will soon be added to support ad hoc queries.
Data Mining. The data mining feature of the beta edition of SQL Server 2005 is used for examining data imported from the data warehouse. ABSi uses two of the nine data mining algorithms included in SQL Server 2005. The data mining server runs the beta edition of SQL Server 2005 on the Microsoft Windows Server 2003, Standard Edition operating system, part of Microsoft Windows Server System integrated server software. It is hosted on an Intel-based server with two 800 megahertz (MHz) processors, 2 gigabytes (GB) of RAM, and 100 GB of extended storage.
The data mining feature of the beta edition of SQL Server 2005 analyzes SMS data to identify cross-sell opportunities. Future plans call for bringing additional data sources into a data warehouse and creating separate data marts for marketing and finance. Bringing additional information into the data warehouse will provide more opportunities for data mining. This should also increase the accuracy of cross-selling predictions. ABSi also plans to upgrade all elements of its data warehouse and data mining solution to SQL Server 2005.
In the future, Project 88 will use data mining to generate the real-time recommendations so when a user places an order for one service, a specific ring tone for example, the system will be able to recommend other tones or related products such as downloadable photographs of the group that created the ring tone.
Benefits
ABSi has enjoyed a number of benefits since deploying the data warehouse and data mining solution. Users now get accurate and up-to-date reports as well as do ad hoc reporting against the data warehouse, gathering information in minutes instead of days, which enables them to make more timely observations and decisions. SQL Server 2005 data mining has helped the company double the response rate to cross-selling promotions. The company has also found that since upgrading to SQL Server 2005, it enjoys eight times faster data mining processing, and faster data mining prediction, allowing them to send out recommendations to more users.
Doubling Response Rates
Moving the recommendation system to SQL Server 2005 has doubled the response rates to ABSi's cross-selling promotions for ring tones. "Our management is very impressed that we could double our response rate through our SQL Server 2005 data mining," says Cunanan. "Every time we present our numbers during management meetings, managers of other services ask us to provide the same magic for them which is what we will do with the full project rollout."
As impressive as the doubling in responses is, Cunanan thinks the data mining recommendations will get even better. "I believe we are only seeing the tip of the iceberg in terms of data mining potential," she says. "We expect to see even greater results as we tune our solution and bring in more data to the warehouse. As we add more information from our other ABSi data sources such as age, music genre of other purchases, and bring in sales patterns from related services, we should see even better predictions."
Cross Selling
Steps 1 through 6 show how data mining provides real-time cross-selling suggestions when customers place an order.
Reporting in Minutes Not Days
SQL Server 2005 Analysis Services is so easy to use that people are obtaining information in minutes that used to take days to do with SQL queries and spreadsheets. "Our users can now generate their own ad hoc reports," says Cunanan. "People are getting responses in a minute or two that used to take days to generate. This makes it much easier to explore market opportunities."
"Project 88 provides managers a platform to analyze timely information so they can respond quickly to the market," says Reyes. "Project 88 has revolutionized how managers analyze their data. They can now look at different views of their data like never before."
Eight Times Faster Data Mining Processing
ABSi initially used the data mining feature of SQL Server 2000. When they upgraded to SQL Server 2005, the data mining team was impressed by how much faster the algorithmic learning process went.
"SQL Server 2005 data mining has a much faster model training process than SQL Server 2000," says Reyes. "The typical models that we use take several hours with SQL Server 2000, but now only take minutes to process with SQL Server 2005 using the same hardware configuration. This huge reduction in learning time will become more important to us as we begin to serve other product groups and run data mining against more data sets."
Faster Data Mining Prediction
The ABSi data mining team was especially happy to see how much faster SQL Server 2005 was compared to SQL Server 2000. "Based on the benchmarks we've collected, SQL Server 2005 is 64 times faster than SQL Server 2000 in processing data mining queries for predictions," says Cunanan. "We need that speed in order to create individualized cross-selling offers for our customers."
The group anticipates the query processing becoming even faster. "In the lab, we are seeing significantly higher throughput achieving more than 1,000 recommendations per second on a four-way server with 4 GB of RAM," says Reyes. "We plan to apply the rich set of algorithms available in SQL Server 2005 in future releases of Project 88 to learn more interesting and useful relationships with ABSi's data, uncover opportunities, and complement their business strategies. SQL Server 2005 brings our data mining capabilities to a whole new level."
Microsoft Windows Server System
Microsoft Windows Server System integrated server infrastructure software is designed to support end-to-end solutions built on Windows Server 2003. It creates an infrastructure based on integrated innovation, Microsoft's holistic approach to building products and solutions that are intrinsically designed to work together and interact seamlessly with other data and applications across your IT environment. This helps you reduce the costs of ongoing operations, deliver a more secure and reliable IT infrastructure, and drive valuable new capabilities for the future growth of your business.
For More Information
For more information about dB Wizards products and services, call +632 7574889 or email dbmarketing@wizardsgroup.com.
For more information about ABS-CBN Interactive Inc, visit http://www.abs-cbnnews.com/.
Microsoft Case Studies - http://www.microsoft.com/casestudies/casestudy.aspx?casestudyid=48615
