A Data Science Central Community
Gilt Groupe is a collection of web sites targeting shoppers with flash sales of merchandise, offering members exclusive discounts of high-end clothing, shoes and other apparel. Once members sign up, they receive e-mail alerts offering sharply discounted designer products targeted based on tastes they’ve indicated from previous purchases. Members have a window of between 36 and 48 hours in which to purchase the products. Product offerings are changed daily.
Gilt originally targeted women shoppers but later added men’s apparel site Park & Bond, as well as the daily deals site Gilt City. To attract shoppers, Gilt offers roughly 30 different sales each day. As a result of the dynamic sale changes, the company keeps inventory light. Whereas most department stores turn over inventory two or three times a year, Gilt turns over inventory eight to 10 times a year. “Inventory can get you into trouble,” Gilt Co-founder and CEO Kevin P. Ryan said. “That can bankrupt you.”
Initially the Gilt information system relies on a highly de-normalized OLTP Data Base, with no reporting infrastructure, slow queries, terabytes of click stream data in flat files, terabytes of data to analyze, high growth context expanding by 2 million rows/day and globally poor performance with query speeds limiting business operations. Gilt needs to get better at personalization. Gilt needs to be able to offer customers different sales than what someone could find in town. Furthermore Gilt wants to realize deeper analysis of integrated customer data, to detect and reacts to fraud, to understand trends in the business through Page View and Clickstream, to continually improve the site experience through analysis of consolidated customer data, and manage its business with relevant statistics and KPIs.
Gilt deploys on Amazon, a 9 node Aster Data installation on AWS. The Aster solution scales horizontally and resources can be added on-demand, supporting continued growth. Aster provides low latency queries over TBs of data. Performance can be tuned by selecting from 11 available AWS instance types (e.g., High-Memory, High-CPU, Clustered for network locality). This solution offers valuable business impacts; in particular the ability to analyze customer/site interaction, a customer support system that personalizes sales, based on browsing and purchase history, and the ability to deploy on Amazon VMs, creating savings over physical machines.
“Within the Teradata Aster platform, we tie together click stream information with email logs, with ad viewing information, with operational information in order to identify what's going on with our customers and how to optimize our marketing spend,” says Geoff Guerdat, director of Data Engineering for Gilt Groupe. “We also leverage Twitter feeds into the Teradata Aster platform to tokenize and parse large volumes of text for sentiment analysis over Twitter feeds. This allows us to be more responsive to the customer and put our finger on the pulse of what's going on.”
If you want to have more information on Big Data Cloud possibility, you can follow the below link to Aster DataBase Cloud Edition solution. Aster Data has partnered with leading cloud infrastructure providers including Amazon Web Services (AWS), Dell, and Terremark to deliver a true analytic platform at cloud scale.