You are here

eXelate Turns Big Data into Smart Data for Marketers with Revolution Analytics and IBM PureData for Analytics


Background

Customer:     

eXelate Inc.
New York, NY (www.exelate.com)

Industry: Ad Tech (Advertising Technology)
Challenge:    Generate hundreds of millions of scores every day using tens of thousands of attributes and terabytes of data compiled from 200+ data sources in order to help marketers target online advertising to individuals with the highest propensity to convert.
Solution: Big Data Analytics solution leveraging eXelate’s proprietary prediction models developed and running on Revolution R Enterprise and IBM PureData System for Analytics, powered by Netezza technology.
Results: Analytics solution performs efficiently at massive scale to support eXelate’s rapid growth. eXelate’s customers realize 3-5 times higher performance (lift) in campaign results, reducing their customer acquisition costs.

eXelate makes big data work for online advertisers.  The company provides third-party data on purchase intent and behavior and the demand for their products is booming.   eXelate’s research found that 68% of agencies, 84% of networks, exchanges and demand side platforms, and 62% of marketers said they would increase budgets for custom data in 2013.  (52% of agencies; 44% of networks, exchanges and demand side platforms; and 43% of marketers already use third-party data.)   Respondents said that custom, first-party data has become the most desirable for branding and direct response campaigns because it helps allocate ad spend smarter, and helps to derive better insights and improves customer targeting.  

By mastering the big data analytics challenges, eXelate has been able to transform its business model into higher value products and services; moving from a company that provides “big” data to a company that provides actionable, predictive “smart” data and information.

Challenge

Because purchase intent and preferences are brand-specific (going beyond category and demographic data), eXelate must build unique models for each of its clients.   In order for eXelate’s output to be timely, the models must be scored several times a day, and the source data must be refreshed just as often. eXelate collects behavioral data from more than 200 partners in its network such as Autobytel, MasterCard Advisors, HomeAway, etc., which amounts to  over 100 terabytes of analytically useful data per year.   Scores are then generated for hundreds of millions of cookies using tens of thousands of attributes. 

eXelate’s challenge is both analytically complex and computationally intense.  Lead data scientist Patrick McCann explains, “Deploying reliable custom prediction functions can be incredibly complicated – we have rather complex prediction functions from training models in R. Deploying these scoring functions in another language can be quite burdensome. For example, our function may apply a tree based algorithm to one subset of users and a regression based algorithm for different subset.   For a different campaign we might apply a completely different classification algorithm, or we can combine many different algorithms into an ensemble where each model gets a weighted vote.  We need a model development and deployment platform that can handle the complexity and the scale of our problem without limiting our toolbox of solutions. Training models within Revolution R and scoring them using native R prediction functions makes that workflow dramatically easier. ”

Solution

eXelate built its big data analytics solution with Revolution R Enterprise and IBM PureSystem for Analytics, powered by Netezza technology (previously called TwinFin).    The Revolution R Enterprise and IBM solution allows eXelate to write custom functions for a classification model and predict who is likely to convert on an advertisement or be positively influenced by brand advertising, and then apply the models to the millions of rows in the database.  “R, the core of Revolution Analytics’ functionality, allows us to deploy complicated prediction functions,” said McCann. 

The Revolution Analytics and IBM solution saves time and money and generates predictive functions that provide superior lift to clients’ marketing efforts.  Because models are built, trained and executed natively in the R language through Revolution R Enterprise, eXelate’s data scientists don’t have to re-write the prediction function in another language for scoring, as is so often the case.    In addition, the company is saving about 50% on analytics infrastructure because it does not have to be replicated to support the model training and scoring process.  Because the infrastructure has been engineered for performance and efficiency, the data science team can iterate through more possible models and get a better result for clients. 

McCann explained, “Combining Revolution Analytics’ capabilities with IBM’s works very well, especially for analytics applications that call for parallel scoring.  Using the nza package’s nzApply & nzTapply functions we can push arbitrary R functions to a database hosted on a IBM PureData System for Analytics appliance.  We know the outcome for millions of users who have seen advertising, and we use their attributes to try to predict the outcome for the hundreds of millions who have not.  We find two advantages to the Revolution Analytics and IBM framework: minimization of data transfer by moving the function to the data as well as the computation speedups associated with distribution of the embarrassingly parallel problem of user scoring.”

“The Revolution Analytics R framework and its seamless integration with solutions like IBM’s Netezza allow for rapid development and deployment of custom-made high performance algorithms for big data analytics,”  explains Matt Fornari, Director of Data Science.

Results

eXelate’s Big Data Analytics capabilities are producing high-value smart data for its clients, and is transforming the organization.  The company generates hundreds of millions of scores per day for all sorts of different products and brands, compiling the preferences of more than 500 million users.  The company is able to quickly build and generate complex analytics for partners so that they can ultimately deliver custom segmentation to client campaigns.  The solution provides efficiency and cost savings in terms of time and infrastructure expense. 

“There are a few important benefits,” explained Kevin Lyons, SVP Analytics.  “It’s really easy to develop and maintain for analysts.   We don’t need a full time DBA to administer the IBM database appliance and don’t need a hardcore programmer to deploy our prediction functions.  We’ve been able to scale our solution to a problem that’s so big that most companies could not address it.  If we had to go with a different solution we wouldn’t be as efficient as we are now.”

He continues, “Customers love it.  They are seeing an increase in the efficiency of their digital marketing campaigns and we are seeing rapid growth in our analytics business.

User

About eXelate

eXelate provides data and insight on online purchase intent, household demographics and behavioral propensities that enable digital advertisers to make optimal marketing decisions. Through the collection of directly measured online data and distribution partnerships with information leaders such as Nielsen, Nielsen Catalina, MasterCard Advisors, Autobytel, Bizo, and more, eXelate makes online, offline, and custom modeled data sets actionable across 500M online consumers worldwide. eXelate’s proprietary maX data™ – customized audiences built for advertisers based on first and third party data – delivers 3-5x better campaign performance as compared to conventional data targeting. As members of the NAI, IAB, Council for Accountable Advertising, OPA and Evidon’s Open Data Partnership, eXelate is a leader in privacy compliant advertising practices.

About Revolution Analytics

Revolution Analytics was founded in 2007 to foster the R community, as well as support the growing needs of commercial users. Our name derives from combining the letter "R" with the word "evolution." It speaks to the ongoing development of the R language from an open-source academic research tool into commercial applications for industrial use.

Though our Revolution R products, we aim to make the power of predictive analytics accessible to every type of user & budget. We provide free and premium software and services that bring high-performance, productivity and ease-of-use to R – enabling statisticians and scientists to derive greater meaning from large sets of critical data in record time.  

We also offer our full-featured production-grade software to the academic community for FREE, in order to support the continued spread of R's popularity to the next generation of analysts.

For customers such as Pfizer, Novartis, Yale Cancer Center, Bank of America and others, our flagship Revolution R Enterprise product stands for faster drug development, reduced time of data analysis, and more powerful and efficient financial models.

About Revolution Analytics

Revolution Analytics was founded in 2007 to foster the R community, as well as support the growing needs of commercial users. Our name derives from combining the letter "R" with the word "evolution." It speaks to the ongoing development of the R language from an open-source academic research tool into commercial applications for industrial use. 

Though our Revolution R products, we aim to make the power of predictive analytics accessible to every type of user & budget. We provide free and premium software and services that bring high-performance, productivity and ease-of-use to R – enabling statisticians and scientists to derive greater meaning from large sets of critical data in record time.  

We also offer our full-featured production-grade software to the academic community for FREE, in order to support the continued spread of R's popularity to the next generation of analysts. 

For customers such as Pfizer, Novartis, Yale Cancer Center, Bank of America and others, our flagship Revolution R Enterprise product stands for faster drug development, reduced time of data analysis, and more powerful and efficient financial models.