Predictive Analytics at Work at eBay, Redbox

by   |   July 23, 2014 5:30 am   |   1 Comments

To understand the myriad ways in which predictive analytics can help business, one has to look no further than two of today’s most prominent retail organizations.

E-commerce giant eBay outlined a forthcoming program for its merchant partners during a presentation at the Predictive Analytics World conference in Chicago, and movie and video game rental company Redbox described how it uses predictive analytics to better understand same-store sales at its 35,000 kiosks located throughout the United States.

Responsible for an estimated $205 billion of ecommerce in 2013 and sitting atop massive amounts of consumer data, eBay is now looking to harness its powerful technical infrastructure to monetize this trove of information, leveraging it for its business partners.

Related Stories

Human Element Improves Predictive Analytics Success.
Read the story »

Redbox finds new kiosk sites by tracking DVD rental behavior data.
Read the story »

How to Get Sales Reps to Adopt and Crave Predictive Analytics.
Read the story »

Use Predictive Analytics to Improve the Customer Experience.
Read the story »

“We understand our customers across a wealth of commerce data,” said Ashok Ramani, Product Lead, Big Data, at eBay.

That understanding includes customers’ payment data (eBay owns PayPal) and data from multiple shopping sites, as well as behavioral data, bidding, and third-party demographics.

Ramani said a unique advantage that eBay enjoys is its high-speed, scalable technical infrastructure—Hadoop clusters and Teradata servers—that can already surface search results in a tenth of a millisecond.

The new program will layer on a predictive model, which can be transferred to a merchant partner without disclosing the eBay customer’s personally identifiable information for use in real-time, targeted content, pricing, and marketing messages.

eBay will transfer only its prediction about the customer’s segment, never personally identifiable information or shopping history. The company will not even disclose the customer’s gender, Ramani emphasized.

For instance, a merchant site can be advised that the arriving shopper belongs to an “adventure-seeker” customer segment—information the merchant can then use to create a dynamic, personalized shopping experience for the shopper.

But eBay’s new program will go further than a one-time transfer of a predicted segment to a partner. It will establish a feedback loop that should improve the prediction model over time.

“When we deploy, we’ll know if the creative caused a conversion,” Ramani said. “That gives feedback and validation [of the model].”

The loop will “train” the model, meaning that eBay won’t need to go back and redesign its predictive models, which will adapt over time. This approach also will let eBay scale the service to many merchants, Ramani said.

The program is being tested in a limited pilot with existing eBay partners beginning this month, running for four to six weeks, Ramani said.

At movie and video game rental company Redbox, the challenge was correctly detecting trends.

The company has experienced tremendous growth in the last few years, and will soon cross the 4 billion movie rental plateau.

But this rapid growth—the number of Redbox’s iconic red kiosks has increased from 5,000 in 2008 to 35,000 today, and are now within a five-minute drive of 65 percent of the U.S. population—can make it tricky to study business and seasonality trends.

Growth was “clouding the data,” said Taly Kanfi, Manager, Strategy and Analytics, at Redbox. Specifically, trends such as same-store sales couldn’t be examined using a simple linear model.

To address this problem, Kanfi began by identifying October 2009 as a turning point in the company’s growth and market share. She then created a model that treated 2008 as if the company had 35,000 kiosks, thus normalizing the data from 2008 onward.

“This let us get value out of our historical data,” she said, adding that, “A little bit of analytical sophistication goes a long way over standard reporting.”

In addition, Redbox uses predictive models to determine which kiosks should be stocked with extra copies of a specific movie title, such the January blockbuster film “Jack Ryan: Shadow Recruit.”

“[We can determine] how many copies of “Jack Ryan” each kiosk gets, and we can leverage that knowledge for the next movie that looks like ‘Jack Ryan,’ ” said Matt James, Senior Director at Redbox.  James said that Redbox bought 400,000 copies of “Jack Ryan” earlier this year.

Ellis Booker is a freelance journalist based in Evanston, Ill. Email him at Follow him on Twitter: @ellisbooker

Tags: , , , , ,

One Comment

  1. Shahriar khan
    Posted November 10, 2014 at 3:15 am | Permalink

    Awesome blog. I enjoyed reading your articles. This is truly a great read for me. I have bookmarked it and I am looking forward to reading new articles. Keep up the good work!

Post a Comment

Your email is never published nor shared. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>