Views from the Marketplace are paid for by advertisers and select partners of MIT Technology Review.
Big Data Game-Changer: Alibaba’s Double 11 Event Raises the Bar for Online Sales
The Double 11 Global Online Shopping Festival, which has taken place annually on November 11 since 2009, has set new standards for online shopping around the world by generating impressive revenues—which are still rising—and creating an innovative retail experience for consumers.
In the just-completed 2016 Double 11 Festival, Alibaba’s shopping platforms achieved a gross merchandise volume (GMV) of $17.8 billion, up from $14.3 billion the previous year. The 2016 event was marked by vigorous online activity, with a peak of 175,000 transactions and 120,000 payments per second. By comparison, the combined online sales for Thanksgiving and “Black Friday”—the annual kickoff to holiday shopping, which occurs annually in the United States on the day after Thanksgiving—totaled about $4.5 billion nationwide in 2015, according to site-tracking data from Adobe Systems.
Alibaba Group, which has evolved from an early e-commerce company into a global technology pioneer, engineered the Double 11 event using big-data algorithms, complex IT architectures, and sophisticated software and data applications. It uses technologies such as artificial intelligence (AI) with machine learning, virtual reality, cloud computing, and mobile Internet, all within a highly secure perimeter. Such technologies drive the company’s online marketplaces and facilitate trading to precisely match buyers and sellers.
Alibaba—whose mission is “to make it easy to do business anywhere”—consists of multiple business units, including e-commerce platforms such as Alibaba.com (a business-to-business e-commerce platform), Taobao (a consumer-to-consumer e-commerce platform), Tmall (a business-to-consumer e-commerce platform), the Cainiao logistics platform, and Alibaba Cloud. The company provides the fundamental technology infrastructure and marketing reach to help companies better engage online with users and customers.
A Big-Data Business
Alibaba focuses relentlessly on big data. In fact, big data underpins all of Alibaba’s efforts to harness the capabilities of everything from cloud computing to AI to smart logistics to digital entertainment. And big data has helped Alibaba build what has become the world’s largest retail platform.
“Alibaba defines itself as a big-data company," says Zhang Jianfeng, the company’s CTO. “To fully realize big data’s benefits, it is essential to lay a strong foundation for managing data quality with advancing data-processing tools and practices that can scale and be leveraged. Alibaba’s data-streaming and processing services have been enabling businesses to discover higher value of their data from diversified dimensions and parameters.”
Zhang cites three key features that contribute to the effectiveness of Alibaba’s data services: first, the data is derived from actual customer shopping actions, making it both more authentic and more useful. Second, the data is highly structured, with more than 100 attributes in discrete units such as color, price, size, and many other parameters, providing better data quality than that offered by a social-network platform. Finally, the data is comprehensive, with many petabytes (PBs) of data streaming in real time from 100 million users per day.
In addition, Alibaba maintains powerful platforms for processing all that data. The company’s cloud batch-data processing platform, MaxCompute, processed 1.98 million computation jobs on the day of the 2016 Double 11 Festival, using more than 54,000 machines in seven data centers (12 clusters) in three locations. More than 180 PB of data was processed that day. Meanwhile, Alibaba’s real-time data-processing platform, StreamCompute, processed a variety of online transaction types and calculated the GMV in real time. Specifically, the platform processed 95.56 million records per second at the peak (a total of 3.7 trillion records for the day).
Alibaba uses innovative AI to build a “smart” business. Personalized search and recommendation engines help e-commerce platforms better "understand" users’ likes and intentions. In addition, the technology can build comprehensive shopper and seller credit systems and valuation models. For instance, robots now handle 95 percent of Double 11’s customer service.
Alibaba’s engineers continue to improve mobile Internet access and usage to tailor content and products to users based on behavioral characteristics. The company carefully studies user behavior data at different shopping stages, explains Gu Xuemei, Alibaba’s vice president and head of the company’s Search Business Unit. From this, the company derives intelligent algorithms to narrow users’ shopping intentions. Then mobile apps push related content and products that help users make purchasing decisions, consequently achieving improved engagement.
“AI technologies have been the key to optimizing the efficiency of our e-commerce platforms,” Gu says.
Alibaba uses various kinds of machine-learning technologies to realize AI, including high-dimensional statistics, online learning, transfer learning, and deep learning. These technologies enable Alibaba to scale its e-commerce platforms to meet customer demand and produce innovative features in its image, video, and speech-recognition technologies.
Overall, these complex technologies work together to facilitate shopping by providing more choices and greater ease of ordering. In e-commerce search alone, applying deep reinforcement learning and online learning would increase GMV by more than 10 percent on Double 11 day.
The strength of Alibaba’s machine-learning models comes from its effective use of billions of data samples and attributes, according to Zhou Jingren, Alibaba’s vice president and chief scientist of Alibaba Cloud. He adds that Alibaba’s server architecture has been developed to handle such large models and their billions of model parameters.
“Through data and model parallelization, we distribute the billions of model parameters into a large set of servers for parallel computing with failover mechanisms”—that is, switching to a standby server, system, or network when failure occurs, Zhou says. “Our most effective machine-learning algorithms have been implemented, and their application produces many benefits.” The sheer scale of data attributes, from 10 million to 1 billion, makes computational advertising work much better, he says. That’s because the approach creates the best match for a particular user in a given context and, as a result, produces a targeted advertisement. Such advertising, of course, greatly helps stimulate sales.
Increasingly, Alibaba has focused on virtual reality (VR), the creation of a virtual interactive world with computer-generated 3-D imaging, and augmented reality (AR), which superimposes a view or image atop the VR view, to provide a realistic, complete, and tailored picture for a specific user.
“We have greatly enhanced our VR and AR ecosystem,” says Alibaba Senior Director Zhuang Zhuoran. “We’re using new hardware and devices to enable it. We’re also pushing upgrades to our shopping ecosystem to enable highly creative 3-D presentations and rich interactions in VR and AR.”
Looking ahead, Alibaba plans to take all these technologies and solutions to precisely match merchant supply with customer demand. That approach is expected to improve user experience, accelerate online sales, and—ultimately—result in happier customers.
Cloud Computing and Cybersecurity Capabilities
To facilitate the enormous volume of online traffic and transactions during the Double 11 Festival, Alibaba leverages cloud-computing technologies such as elastic computing, virtualization, and real-time data processing backed by the world’s largest hybrid-cloud architecture.
Apsara, Alibaba Cloud’s super computational engine, “ensured that networks and systems were scaled to process during Double 11, especially during the unpredictable traffic spikes, without incident,” says Alibaba Senior Director Wu Zeming. Alibaba’s scale-fault tolerance—the ability to provide backup with seamless operation upon a fault—also results in a flexible and robust infrastructure for all platforms.
Alibaba Senior Staff Engineer Tang Zinan says that the company’s cloud-computing platform is also strategically built to allow small and midsize businesses to operate more efficiently. Cloud computing provides flexibility based on demand, providing advantages both for Alibaba’s internal operations and for its many external partners and enterprise customers.
At the same time, Alibaba is focused on tightening up cybersecurity for all shopping transactions. “We use high security standards for the whole data lifecycle: from data collection, use, transfer, and sharing, to data disposal. We also have promised our customers that we do not to touch or share individual data,” says Zhang, the company’s CTO. “Our security measures use sophisticated technologies for data sorting, classification, and authorization management, as well as access risk monitoring and early warning. The intention is to have a comprehensive approach to protecting our customers’ privacy and their data within our ecosystem.”
And a vast ecosystem it is, including 10 million merchants, tens of thousands of vendors, and more than 3,000 logistics companies—all providing products and services to customers within a highly secure environment. The security strategy in this far-reaching environment includes capabilities for early detection of cyberattacks and abnormal behavior as well as automated risk-case tracking.
Moving Toward “the Infrastructure for Global Commerce”
Since its inception as part of China’s “Singles Day,” a nationwide holiday intended to help young Chinese people celebrate being single, the Double 11 Global Online Shopping Festival continues to grow and set a high bar for the online customer experience.
But while Alibaba will continue establishing a preeminent global shopping presence, the company is striving to go beyond online sales. “E-commerce is just the first phase of Alibaba’s entire strategy,” says Zhang, the company’s CTO, adding that half of Alibaba’s employees already work in areas such as big data, cloud computing, AI, mobile Internet, smart logistics, and digital entertainment.
“Our ultimate strategy is to build the future infrastructure for global commerce,” he says. That’s an undertaking that, while highly ambitious, remains true to Alibaba’s core mission of making it easier for consumers and companies alike to do business anywhere.
Become an MIT Technology Review Insider for in-depth analysis and unparalleled perspective.Subscribe today