Web-only retailers, including Amazon, accounted for 42% of sales of all retailers ranked in the Read Now
Mupd8 software from @WalmartLabs channels “big data” to enhance marketing, Wal-Mart says.
Some topics are so hot online they generate thousands of Twitter tweets per second. Extracting a useful signal out of that noise and responding immediately with, say, a product recommendation based on the tweets a consumer is following, requires the capacity not only to analyze that large volume of incoming data, but to do it fast. Wal-Mart Stores Inc., No. 4 in the Internet Retailer Top 500 Guide, through its research division @WalmartLabs has developed a new computing platform with that capability, called Mupd8, and is already using it to improve online marketing, the retailer says.
For example, Mupd8 (pronounced mup-date ) powers Wal-Mart’s Shopycat Facebook application, which analyzes within seconds all the Facebook status updates, Likes and interests of a consumer’s friends to suggest gifts for her to buy for them says STS Prasad, @WalmartLabs’ vice president of engineering. The platform is meant for use “where you want to leverage events as they happen to improve customer experience,” he says.
The retailer also uses Mupd8 to collect video, images, location information and status updates from social media streams to determine which trending products it should carry in stores or feature in online marketing, Wal-Mart says. For example, the retailer noticed that social chatter about Takis tortilla chips was growing in some states, which influenced its decision to stock those chips in nearby stores, it says. So far Wal-Mart is using Mupd8 in more than a dozen projects, processing more than 300 million status updates per day, it says.
Mupd8 could change the way e-commerce sites operate, Prasad says. A Mupd8 application could, for instance track web site activity, detect when visitors are encountering a problem and in an automated way suggest what they should, reducing customer service calls and e-mails, Wal-Mart says.
Mupd8 by design requires no new data centers or specialized hardware to run, making it cost-effective and scalable, Prasad says. He says that’s possible because the software splits the processing tasks among several machines.
It is based on similar systems developed over the last decade to handle big data, he says, such as the open source Apache project Hadoop which began at Yahoo Inc. Open source refers to computer code that is publicly available on the Internet for developers around the world to tweak and use to build applications. When they do, most typically share at least the base code for what they’ve created, generating a collective library and helping the platform to evolve at a faster pace than it might if the code were kept secret by the company that developed the software. Last month, Wal-Mart made Mupd8 open source.
“It’s one of the first times Wal-Mart has open-sourced a technology infrastructure of that scale,” Prasad says. “Wal-Mart leverages open source [technologies] in many areas in the e-commerce site and it’s only fair we contribute back.”
The platform’s design is based on a big data analytics framework similar to Hadoop called MapUpdate, he says, to which it partially attributes its name (@WalmartLabs formerly referred to Mupd8 as Muppet). It derives from work at social media technology company Kosmix, which Wal-Mart acquired about a year and a half ago. There, engineers were looking for ways to identify breaking news by analyzing large numbers of tweets per second, Prasad says.
By making Mupd8 freely available, Wal-Mart hopes that other creative minds will come up with even more e-commerce applications for quickly processing enormous amounts of data, he says. “Technology projects tend to benefit from having an open source system that in many senses can take it beyond where any one company can,” he says. Prasad adds, however, that he expects it to take a year or more before Mupd8 developers start producing e-commerce applications that other retailers may want to adopt.