How e-commerce giants are using AI and marketing (Part 1)
In the first part of a series, columnist Daniel Faggella explores how e-commerce leaders such as Amazon and Alibaba are adopting AI applications to power more accurate product recommendations and faster search results.
Groceries, toys and flowers are no longer the only things filling UK-based Ocado’s massive Hampshire warehouse. In the two years since the company started branding itself as the world’s largest online-only supermarket, the distribution center has begun teeming with robots steered by algorithms that control the logistics side of the company’s 24/7 grind. Now, frozen meat and canned tomatoes are sorted and picked up by robots, which allows the customer service team to deliver them to the doorstep more quickly.
The explosion of AI in e-commerce has allowed online retail giants to explore advanced technologies. As firms in the e-commerce industry continue to expand and cater to a growing number of customers, the need for an automated system to streamline their operations has become one of their top investment priorities.
Ocado reportedly planned to spend £175 million ($232 million) for its robotics and automation system this year, Amazon committed $5 billion in its tech investments in India alone, and Alibaba is looking into pouring $15 billion into research and development spending over the next three years. Forrester Research (PDF) predicts that AI-related investments will grow by about 300 percent in 2017, with businesses becoming more competitive by 2020 and gaining $1.2 trillion per year.
But current technology is far from perfect. In the case of Ocado, it failed to impress analysts during their visit to the company’s Andover facility. According to a Telegraph interview with George Mensah of Shore Capital:
The slightest error in terms of the movement of the robot will cause them all to pause, so there were periods where there were long pauses.
The consensus was that the technology still had obvious glitches that needed a lot of fixing, which stands in contrast to the flawless operation Ocado shows on its marketing video.
To better gauge the impact of AI-powered interactions by e-commerce giants’ tech tools, we researched this field in depth to help answer questions such as:
- What AI applications are adopted by leading e-commerce companies?
- What impacts have these AI applications contributed to these companies’ operations?
- What AI use patterns do these companies have in common?
This two-part article presents a comprehensive look at some of the leading e-commerce firms and their adoption of AI in their operations. The firms were selected according to their sales revenue rankings in 2016.
Artificial intelligence in e-commerce
The most popular AI applications from the selected e-commerce companies using AI in their marketing operations include:
- chatbots to improve customer service.
- image and voice recognition for faster search results.
- recommendation engines with advanced algorithms for more accurate product recommendations.
In this article, I’ll explore how AI is implemented in each e-commerce company.
Before Ocado had its robots stuffing orange plastic bins and running them on winding conveyor belts, Amazon was looking to get ahead of its competitors by forming Amazon Robotics after purchasing Kiva Systems in 2012. The subsidiary is continuing to develop robotic technology using machine learning, object recognition and computer vision in Amazon’s fulfillment centers.
– Product recommendations
If Amazon’s latest earnings are any indication, product recommendations powered by AI also have been successful. The company reported a third-quarter sales increase of 34 percent to $43.7 billion. The recommendation system is integrated into every aspect of the purchasing process. (The company, however, prefers not to disclose how effective its system is.)
But a team of University of Toronto professors notes in a Harvard Business Review article that Amazon’s system still doesn’t deliver 100% accuracy in its predictions. They wrote:
Amazon’s AI does a reasonable job, considering the millions of items on offer. However, they are far from perfect. In our case, the AI accurately predicts what we want to buy about 5 percent of the time. In other words, we actually purchase about one out of every 20 items it recommends.
That said, the team went on to predict that given additional data (such as that provided by Amazon’s purchase of Whole Foods), the company could eventually become so accurate that it could someday turn a profit by shipping people items it predicts they will need.
– Battling fake reviews
Amazon’s well-known product reviews can help in marketing, but some companies have found a way to generate fake feedback to boost their product’s ratings on the site. Sensing the growing number of inauthentic reviews on its website, Amazon filed a lawsuit in 2015 against companies that have paid for positive feedback, a scheme known as “crowdturfing.” Some have also paid for negative reviews to be posted against their competitors. To combat the proliferation of fake reviews, Amazon released a machine-learning algorithm to better filter authentic online feedback.
However, current developments in technology may just bolster the fake reviews business. Now, instead of companies paying people to post positive reviews, they may tap AI to do it for them. A team of researchers (PDF) from the University of Chicago used a neural network to prove that it can write phony positive reviews that are indistinguishable from human-written reviews, as judged by the humans who verified the content. Their research has since been featured by multiple news outlets to raise awareness of the misuse of AI.
– Style recommendations
The fashion industry is where Amazon is planning to use AI to improve its marketing reach. In April 2017, it launched Echo Look, a hands-free camera assistant and personal stylist which uses a combination of human advice and machine learning. Users can command Alexa, Amazon’s cloud-based voice service assistant, to take a photo or video of their outfit to be posted on social media. It also can compare outfits using Style Check’s data analytics.
The Wall Street Journal’s Geoffrey A. Fowler put the technology to the test by inviting stylists and then comparing their taste versus Alexa’s. He was surprised that Alexa was mostly on point, although not always.
Amazon continues to invest in AI for a variety of reasons. As CEO Jeff Bezos wrote in his annual shareholder letter, “Machine learning drives our algorithms for demand forecasting, product search ranking, product and deals recommendations, merchandising placements, fraud detection, translations, and much more.”
Rajeev Rastogi, the company’s Machine Learning lead, told the Economic Times, “We are applying AI to a number of problems such as speech recognition, natural language understanding, question answering, dialog systems, product recommendations, product search, forecasting future product demand, among others.”
Amazon Web Services, the company’s cloud-computing platform subsidiary, has developed these plans in the form of deep-learning tools such as Amazon Lex (a chatbot service for building conversation interfaces), Amazon Polly (a cloud service that turns text into speech) and Amazon Rekognition (a deep-learning image analysis technology). Amazon also has teamed up with Microsoft to develop Gluon, a user-friendly interface for developers working on training algorithms and neural network models.
Alibaba, one of China’s e-commerce behemoths, has been investing in technology to improve its services, especially in marketing.
– Product recommendation
Its own software, E-commerce Brain, powers the company’s product-recommendation technology. It builds predictive models using real-time online data on content consumption, buying behavior and other data from the entire Alibaba ecosystem such as Alipay, AutoNavi, Youku and UCWeb.
The company’s AI also aids sellers by helping them create product-buyer matches on personalized virtual storefronts to improve the chances of selling their products. The recommendations are based on buyers’ purchase history, background and location, among others. Alibaba claims that through this technology, they recorded a reported 20 percent increase in conversion rate during their 24-hour online shopping event in 2016.
The company also launched Dian Xiaomi (or store assistant), another AI-powered text-only chatbot, earlier this year to help merchants customize and manage their virtual storefronts, especially during rush hours when they are short of human staff to entertain buyer inquiries and problems.
– Smart supply chain
In addition, the company is exploring the development of a smart supply chain in China through its Ali Smart Supply Chain (ASSC) platform, which predicts volatile buyer trends so sellers can focus on improving their product, inventory and delivery operations. Supply chain design can spur innovation and improved production, along with improving infrastructure through smart cities. After developing a smart transportation system for Hangzhou in eastern China, Alibaba has now turned to building an AI hub in Macau over a four-year period, bringing advanced technology to transportation, health care and governance.
In 2015, Alibaba claimed to have pioneered China’s first AI platform when it launched DT PAI. The cloud offering allows companies to use tons of data and analyze which target market suits their products well. During its development phase, the company was planning to use image recognition by letting users take a photo of an item and allowing the platform to direct them to a specific page on the website where they can purchase it.
In an interview (PDF) with MIT Technology Review, Felix Liu, Alibaba’s head of Customer Experience Business Group, explained that AI has helped the company by obtaining massive customer data, discovering key issues and improving the company’s inquiry management channels. In one instance, Liu cited how their chatbot was able to detect abnormal levels of order status inquiries and took only 30 minutes to provide a solution to their affected customers. As Liu put it, “AI currently helps 100 percent of our customers with inquiries, and resolves 50 percent of them completely.”
In October 2017, CTO Jeff Zhang announced during their computing conference that the company will be investing $15 billion in research and development over the next three years, much more than the investments they made from 2014 to 2017. They will build seven research laboratories, dubbed the Academy for Discovery, Adventure, Momentum and Outlook, spread across different continents. Researchers will work on projects dealing with AI, machine learning, natural language processing and data intelligence, to name a few.
(In Part 2, I’ll cover eBay, Rakuten, Flipkart and ASOS.)
Opinions expressed in this article are those of the guest author and not necessarily MarTech Today. Staff authors are listed here.