Published on 15-Jun-2011
Validated on 03 Dec 2012
"The move to cloud has really helped us because we can turn on these systems much faster and more cost effectively than ever before." - Brian Lent, founder and chief technology officer, Medio Systems
IT/infrastructure, Cloud Computing, Customer Relationship Management, High Availability , Linux, Optimizing IT
Based in Seattle, Medio Systems is a leader in the emerging field of predictive analytics, quickly analyzing data streams to give mobile phone carriers and providers actionable insight into their customer’s preferences and buying habits.
Medio needed a way to process the massive amount of incoming data with a solution that was affordable and scalable.
Medio Systems worked with IBM to deploy more than 60 IBM® System x® 3650, 3550 and 3250 servers featuring Intel® Xeon® multicore processors at two data centers. The implementation included Red Hat Enterprise Linux and Apache HBase Hadoop software to power the company’s data-intensive distributed analytics infrastructure. The service was then deployed as a cloud-based platform architecture.
The IBM solution slashes time to process customer responses by a factor of 12, boosts the number of transactions per second by 30-fold and offers 100 percent system availability.
In a saturated mobile phone market where retaining existing customers has become at least as important as adding new ones, understanding user likes and dislikes has always been a top priority to fuel business growth. But until recently, gaining this insight was a cumbersome, costly process.
What if you could accurately anticipate customer preferences while saving money and time? An emerging mobile targeted marketing technique using predictive analytics is doing just that. Instead of creating reports based on database queries, which then get used by call centers to contact customers and learn more about their preferences, predictive analytics can process data streams in near real time. This helps businesses deliver to customers uniquely personalized content, applications and promotions on mobile devices.
Based in Seattle, Medio Systems is at the forefront of the predictive analytics revolution. With 65 employees, the startup offers a suite of predictive analytics services for major cell phone carriers who represent a marketplace of more than 75 million customers (over 2 million of which are seen daily) and over 7,000 different types of phones. The resulting metrics give Medio clients the insight they need to provide customers with more relevant content, offers, and experiences.
Adding value with predictive analytics
Compared with traditional customer relationship management techniques, predictive analytics is cheaper, faster and far more effective at reducing a customer churn rate which averages about 1.5 percent monthly in the mobile phone business. With predictive analytics, customers who receive personalized information and offers through their cell phones are 300 to 400 percent more likely to stay with their existing cell phone provider, says Rob Lilleness, chief executive officer at Medio Systems.
“Predictive analytics is really focused on taking similar data and providing fiscal analysis to understand what users might do in the future versus only looking at historical data,” says Lilleness.
Medio initially had a traditional server management-style infrastructure model where co-located servers were deployed for individual clients to process customer data, explains Damon Lanphear, general manager and chief analytics architect at Medio Systems. However, the exploding amount of user data being generated by Medio’s clients, coupled with bandwidth constraints and the need to deliver analytics in as close to real time as possible, was more than this model could support.
“When we’re looking at the growth of mobile data, we’re talking on the order of hundreds of terabytes,” says Lanphear. “One customer told us they generate one to 10 petabytes per year.”
Enabling business intelligence with IBM System x
When Medio began investigating how it could best put predictive analytics to work for its clients and their customers, a number of complex challenges became readily apparent. Beyond the massive amount of storage and processing power required, Medio had to be able to support rigorous service level agreements, including one stipulating “five nines” or near 100 percent availability. Because the standard service delivery model had grown overly complex and costly with the explosion of user data, Medio also wanted to simplify its business model by offering its predictive analytics product to customers as a cloud service. The solution enables Medio to effectively deploy analytics cloud-based services for mobile data monetization.
“The move to cloud has really helped us because we can turn on these systems much faster and more cost effectively than ever before,” says Brian Lent, founder and chief technology officer at Medio. “We can focus on what’s best for our consumer rather than worry about how to build the perfect infrastructure.”
The new solution is anchored by fully redundant, multi-tiered data centers in Seattle and Atlanta. Split between these two locations is a production environment with more than 60 IBM System x® 3650, 3550 and 3250 class servers. The servers feature Intel Xeon multicore processors, which address the company’s need for high-performance analysis of large data sets. Together, IBM and Intel have provided a best-of-breed solution that delivers peace of mind.
Medio’s analytics software runs on Red Hat Enterprise Linux, and the Apache HBase Hadoop database underpins this data-intensive distributed analytics infrastructure. To deliver these services to customers, Medio worked with IBM to deploy a cloud-based platform architecture that was purpose-built for the wireless industry to handle rapidly scaling demand.
“There are a lot of redundancies we needed built in and there was a lot of support behind us in order to make it successful,” says Michael McMillan, director of network operations at Medio Systems. “And that really was something that IBM had proven themselves with.”
Substantial benefits fuel the bottom line
The benefits of the overall solution have been staggering for a service which currently handles more than 2 million daily unique visitors and more than 10 million unique visitors per month. Revenue has tripled in the past year, and to date, Medio’s predictive analytics platforms have supported more than 75 million unique users who have generated 5.5 billion personalized recommendations.
The time required to process customer responses fell with the new solution by factor of 12, from 894 milliseconds down to 75 milliseconds. Along with this speedup, the number of transactions per second increased from 41.5 to 1,200 per datacenter—a 30-fold improvement. And since the upgrade, Medio has maintained 100 percent availability.
“With our new infrastructure, we can deliver products and services faster and more affordably,” says Lent. “IBM gave us that dynamic element so that a small company can grow to meet big company needs quickly.”
Driving forward with IBM intellectual horsepower
Medio Systems tapped IBM’s expertise in big data to ensure that the overall solution delivered the desired performance characteristics at the right price. Now that Medio’s predictive analytics platform can scale to handle tens of millions of potential users, company executives say they are looking at expanding their offerings into overseas markets in addition to Canada and the United States.
“To have the intellectual horsepower of IBM come in and sanity check our network architecture, the type of products that we’re using and the individual nodes of that network has been very important,” says McMillan. “Working with IBM has been amazing and the results have been phenomenal.”
For more information
Contact your IBM representative or IBM Business Partner. Visit us at: ibm.com/systems/x
For more information about Medio Systems, visit: www.medio.com
Products and services used
IBM products and services that were used in this case study.
System x: System x3250 M2, System x: System x3550 M2, System x: System x3650 M2
© Copyright IBM Corporation 2011 IBM Systems and Technology Group Route 100 Somers, New York 10589 U.S.A. Produced in the United States of America June 2011 All Rights Reserved IBM, the IBM logo, ibm.com, and System x are trademarks or registered trademarks of International Business Machines Corporation in the United States, other countries, or both. If these and other IBM trademarked terms are marked on their first occurrence in this information with a trademark symbol (® or ™), these symbols indicate U.S. registered or common law trademarks owned by IBM at the time this information was published. Such trademarks may also be registered or common law trademarks in other countries. A current list of IBM trademarks is available on the web at “Copyright and trademark information” at ibm.com/legal/copytrade.shtml Intel, the Intel logo, and Xeon are trademarks of Intel Corporation in the U.S. and/or other countries. Linux is a registered trademark of Linus Torvalds in the United States, other countries, or both. Other company, product and service names may be trademarks or service marks of others.