When GEO decided to retire its existing data warehousing infrastructure, the marketing team looked to move data analysis to the cloud—but its first-choice cloud solution proved to be slow and difficult to use.
Migrating to IBM Db2 Warehouse on Cloud provided a platform that minimized friction for users by being highly compatible with existing Oracle SQL code, and easy to integrate with SAP Predictive Analytics applications.
75% reductionin excess coupon distribution, saving approximately USD 32,000 per week
9 millioncustomers now receive personalized recommendations through association analysis
Drivesmachine learning and deep learning to reveal insights into recommendation data and increase sales performance
Business challenge story
Finding a new data warehouse platform in the cloud
Founded in 1986, GEO has grown from a single video rental store in Aichi prefecture to a nationwide network of more than 1,800 stores, and has expanded its GEO Shop brand to encompass the rental and sale of new and second-hand movies, videogames, books and comics.
The company owes its success to its outstanding marketing power. The company’s GEO Shops record more than 100 million new transactions per month, and the company performs ad hoc analysis of 36 months of sales data to support rapid marketing decision-making.
In 2011, GEO established a fully-fledged data-driven marketing system, using an Oracle Exadata data warehouse appliance to accelerate the aggregation of sales data for marketing analyses. However, in 2014, when the company’s marketing strategy was starting to take off, GEO made a major change in corporate policy that required some blue-sky thinking from the marketing team.
Mr. Takeo Suzuki from the Media Store Planning Department of GEO’s Media Division recalls: “A decision was made to retire the Oracle Exadata appliance as soon as its maintenance contract expired. The system would be replaced with a new database that would specialize in supporting the core business processes of the entire group. Other departments were advised to procure their own separate data warehouses to support marketing analytics and data science.”
GEO looked for a cloud-based data warehouse service, and initially decided to introduce Amazon Redshift, running on Amazon Web Services—a popular platform at the time. However, the transition to this new platform was not a success.
Mr. Suzuki comments: “We had set up our existing database using Japanese names for the tables and fields. But at that time, Amazon Redshift did not support multi-byte characters in database field names. Also, because we signed up for the pay-as-you-go model instead of a dedicated instance, it took more than 30 minutes to spin up a new instance and load all of our data. We couldn’t satisfy the essential requirements for ad hoc analysis—that when you think of a question, you want to get the answer right away. So gradually we phased out the use of Redshift.”
Capitalizing on existing know-how
Determined not to make the same mistake twice, GEO decided to move to a different cloud data warehouse platform: IBM Db2 Warehouse on Cloud.
“The decisive factor was that Db2 Warehouse on Cloud supported SAP Predictive Analytics—the same analytics software that we had been using with our old Oracle Exadata platform,” says Mr. Suzuki. “It was also highly compatible with Oracle SQL. In fact, the IBM solution can even execute unique Oracle SQL functions such as NVL and ADD_MONTHS. So it allows us to make good use of the know-how around ad hoc analysis that our team has already built up—which is very satisfying.”
In other words, IBM Db2 Warehouse on Cloud's powerful Oracle compatibility features eliminated the need for GEO to rewrite applications that had originally been written for its previous Oracle database. This helped to accelerate the adoption of the new cloud database platform.
There was just one concern: IBM Db2 Warehouse on Cloud had only recently been released, and there was a lack of textbooks and documentation available about the solution in Japan. GEO was initially concerned about the lack of information about how to tune database performance or recover in the event of a problem.
Fortunately, the IBM support team was able to dispel these worries. Mr. Suzuki explains: “I thought that cloud services were to be used at the user’s own risk. But with IBM, the experience was quite different—we received in-depth support from the IBM team.
“Moreover, because Db2 Warehouse on Cloud has an architecture that does not require tuning, we were able to get the performance we expected from day one.”
In parallel with the implementation of the new database infrastructure, GEO’s Media Division also decided to strengthen its ad hoc analysis capabilities. All eight members of the new Media Store Planning Department are being trained to be self-sufficient in using the software for data-driven planning and analysis work.
Mr. Suzuki comments: “The training started with basic SQL skills, because most of the team are from the line-of-business side, rather than the IT side. The first task I set was for everyone to learn how to do ABC analysis using SQL window functions. Steadily, everyone completed the task and became more comfortable with the system.
“Even though their database skills are relatively immature, each member of the team has a wealth of business experience and knowledge, so they are highly sensitive to what the data means. They can also clearly imagine the benefits that analytics will bring, so they are learning fast, and we’re already beginning to use their analyses in our business metrics.”
He adds: “This is the real thrill of line-of-business-led analytics. Db2 Warehouse on Cloud supports these efforts, because as a cloud database, we can use it freely without having any impact on the work of other departments.”
Optimizing marketing strategies with analytics
IBM Db2 Warehouse on Cloud is already generating data that GEO uses in various marketing metrics. For example, the solution has been used to optimize the distribution of coupons that the retailer uses to encourage customers to continue renting comic books.
“Until recently, we used a very simple algorithm to narrow down the groups of customers we wanted to target with coupon distribution,” says Mr. Suzuki. “But when we did an A/B test, we found that we were targeting people who would have visited even if we hadn’t offered them a coupon.
“We created a regression model that calculates the probability that a customer is going to visit, based on the customer’s attributes and usage patterns. As a result, we were able to reduce the excess distribution of coupons by 75 percent, by ensuring that we only target customers who have a low probability of returning.
“Coupon optimization has delivered cost savings of around JPY 3.5 million [USD 32,000] per week. By using this saving as a source of funds, we have been able to return the benefits to a wider range of customers.”
GEO is also using the platform to improve its delivery of personalized recommendations via email newsletters and smartphone apps.
“We have tried a variety of pre-built recommendation engines, but they are difficult to tune to our requirements,” says Mr. Suzuki. “So we built our own association analysis logic in SQL, and we have implemented it using IBM Db2 Warehouse on Cloud. As a result, we can generate weekly recommendations for nine million customers who have rented DVDs or CDs from us over the past six months—with high accuracy, and delivered at pace.”
For item-specific recommendation data, GEO uses open source tools such as Jupyter Notebooks and Cytoscape, which use Python libraries to build and execute machine learning and deep learning models. Each store uses these models as reference data to help improve its product mix and shopping experience.
“Instead of supplying the information to users via an application, we are distributing the Python code itself to our business users,” says Mr. Suzuki. “The aim is to provide an opportunity for each of our analysts to start exploring machine learning and deep learning with Db2 Warehouse on Cloud.”
With the rise of competitors such as video on demand (VOD) services, it is especially important for GEO to continue to innovate. “In the battle for market share, it is vital for us to raise our internal literacy around ad hoc analysis, so that we can strengthen our competitiveness for the future,” says Mr. Suzuki.
For example, GEO is now expanding its own digital marketing platform, and is working on improving the in-store experience for customers by adopting RFID and Internet of Things technologies. By thinking outside the box and investing in cross-functional analytics, the company hopes to be able to anticipate changes in customer preferences and build new businesses accordingly.
With a store network that stretches from Hokkaido to Okinawa, GEO Corporation operates a number of retail brands including GEO Shop for movie and music rental, as well as the purchase and sale of videogames; Second Street for the purchase and sale of clothing, furniture and home appliances; and GEO Mobile for the purchase and sale of second-hand mobile phones and smartphones.
Take the next step
IBM Analytics offers one of the world's deepest and broadest analytics platform, domain and industry solutions that deliver new value to businesses, governments and individuals. For more information about how IBM Analytics helps to transform industries and professions with data, visit ibm.com/analytics. Follow us on Twitter at @IBMAnalytics, on our blog at ibmbigdatahub.com and join the conversation #IBMAnalytics.