Deep Data Intelligence Draws TradeDoubler Closer to Customers
Turning transactions into useful information is no small task when you process 20 billion online transactions every month.
TradeDoubler processes and analyzes data from 20 billion online transactions each month, providing customers such as Dell and Apple Computer with the information they need to optimize their Web marketing campaigns across Europe and Asia.
The company wanted to add some critical business intelligence (BI) capabilities to its cross-marketing media platform that could make its services even more valuable to customers. TradeDoubler went on a mission to find a vendor partner that would provide the ability to mine much more information from its database to answer client questions faster and provide a more complete, insightful picture of the effectiveness of marketing campaigns.
The Challenge
TradeDoubler's existing data warehouse wasn't designed to handle complex analytic queries. When handling anything complicated, TradeDoubler had to carry out a time-consuming retooling and re-indexing of the database. "We had one person working with the data full time, and depending on the complexity of the queries, it took anywhere from half a day to two days to get the data out," said TradeDoubler's chief technology officer Ola Uden.
After deciding to make a change, TradeDoubler evaluated Infobright's analytic data warehouse solution.
TradeDoubler turned to Infobright because, unlike other systems, its solution doesn't require data partitioning or indexing, which significantly reduces the time and effort for implementation and ongoing administration. Instead, Infobright combines a column-oriented database, ideally suited for analytics, with its Knowledge Grid architecture to achieve superior performance without the work required by other data warehousing systems.
TradeDoubler loaded Infobright on a standard Dell server, proving the Infobright product was ready to install off the shelf. Data was extracted from TradeDoubler's Oracle database. More than 3.2 billion rows of data were loaded into the Infobright analytic warehouse at an impressive average rate of more than 300,000 rows per second. Though data compression rates varied by data type, Infobright compressed fact tables at a ratio of 39:1, using algorithms tailored to specific data types.
A similar test on an Oracle 9 Enterprise Edition database (with partition option) running on a larger server was shut down after about two hours because the trial was taking too much time.
"Infobright was fastest, hands down," said Mats Johansson, a senior consultant at Lincube Group AB, who helped TradeDoubler with the implementation. "It handled huge volumes of data and resolved queries faster than I have ever witnessed after years in this industry -- and on standard hardware."
Measuring the ROI
TradeDoubler began using Infobright in May, running it on a $12,500 Dell server with two quad-core processors. The results are impressive.
Before, the Oracle database couldn't quickly produce the detailed statistics and business analysis clients needed. In some cases, queries couldn't be resolved at all. Now, TradeDoubler loads and rebuilds the database every day, retaining three days of network-wide clickstream data and 60 days' worth of online order information. The database can load 3.2 billion rows of data at an average rate of 300,000 rows per second. Certain queries that didn't return at all using Oracle are resolved in less than a minute on the Infobright system, and other queries that took as much as 15 minutes using Oracle are returned in 22 seconds with Infobright.
Additionally, it's also much easier today for TradeDoubler to track user behavior through deeper analysis of Web clicks, impressions, and purchases. Infobright enables TradeDoubler to drill down, construct, and reconstruct hundreds of data points per interaction to get a better understanding of what motivates a customer at a given moment. Using this information, companies can easily identify buying patterns, analyze how many users abandon the site at a particular point, or how an ad campaign fared in 2008 compared to 2007, filtering results by date, type of sale, or product category.
"Infobright helps us to analyze and present a large amount of data that we couldn't analyze before," Uden said. ""Suppose Apple launches a campaign for the iPhone and the company wants to understand how people ended up buying one at their site. We can now tell them what customers did before they made a purchase – whether they read a review first or were responding to an ad. We know which sites they visited before they ended up at the Apple store."
Infobright's quicker loading speeds, automated indexing, 39X data compression, and faster query times combined are providing TradeDoubler with fast, comprehensive answers for clients at a lower cost than would have been possible with conventional technology. From this base, TradeDoubler expects to continue developing its business intelligence solution by adding new features and making it even easier to use in the coming months so clients will gain an even deeper understanding of their Web traffic and how even the smallest change impacts online behavior.
The end result: a more complete picture of the entire customer journey.