Business Information

Technology insights for the data-driven enterprise

Guido Vrola - Fotolia

Preparing data for analysis helps JTV find predictive gold

Predictive models help Jewelry Television's on-air hosts sell its wares, thanks to data integration and preparation processes that funnel a mix of data into the analytics applications.

Jewelry Television, a shop-from-home network and online retailer known as JTV, is moving to add some analytics sparkle that goes beyond after-the-fact reports. It now puts information from predictive analytics models on the screens of on-air hosts to help them better align their jewelry and gemstone pitches with viewer demand.

But the new focus on predictive modeling is also pushing the Knoxville, Tenn., company to polish up its approach to integrating and preparing data for analysis.

Regression analysis algorithms that aim to identify relationships between different data variables are among the types of predictive models being used at JTV, according to CTO Chris Meystrik. The advanced analytics applications drive real-time business operations and decisions; for example, their output shows up on displays in the broadcast studio that look like "an airplane cockpit," Meystrik said.

Getting predictive information on customer buying patterns in front of JTV's hosts helps them focus on selling items that are on the way up the curve of fashion popularity. At the same time, they can use analytics data to avoid overhyping items that may be available only in small quantities, a sales misstep that could disappoint customers interested in those goods.

To make the predictive modeling possible, the shopping network uses data integration tools from Informatica to move a mix of customer relationship management, inventory, order data and other information into its analytics systems. Meystrik said a notable difference from the past is that much of the required data tends not to already reside in a data warehouse -- social media and website clickstream data, for example.

There are also differences in data preparation between basic reporting and predictive analytics. In Meystrik's view, the initial steps required in preparing data for analysis can be quite cumbersome. "But once we have the algorithms on a solid footing and operating with acceptable precision, they generally become self-sufficient," he noted.

Yet things could get even more complicated. To push its operational analytics efforts forward, JTV is currently experimenting with machine learning and open source data streaming and message queuing software. As jewelry sales move further into a post-brick-and-mortar era, network executives see analytics applications as a competitive differentiator for the company. 

Article 6 of 10

Next Steps

Discover the value of faster techniques for building predictive models

Cloud systems increase workloads on preparing data for analysis

Sit in on an online chat covering predictive and prescriptive modeling

Dig Deeper on Enterprise data integration (EDI) software

Get More Business Information

Access to all of our back issues View All