AI in retail: It’s just the beginning

We noticed that you bought three pairs of shoes at this time last year and have set aside a few new designs that you can consider for your Ramadan outfits. Did we also mention that there’s a special discount on items from your favourite footwear brand’s 2019 spring-summer collection?

With artificial intelligence (AI) emerging as the engagement ‘engine’ and data from shoppers acting as the ‘fuel’, conversations between retailers and shoppers are starting to look a lot like the above. In fact, retailers like Shopee and Zalora are already deploying machine learning—a subset of AI—to analyse shoppers’ browsing and purchasing behaviour, before recommending items of relevant interest, all in real-time.

Almost two-third of organisations in the Asia-Pacific region indicate a “perceived need” for AI[1], so it is safe to say that the technology and its proven use cases are ready for primetime. But why are retailers in the region still not fully on board the AI bandwagon?

Online retailers getting physical

Asia is the frontrunner in retail growth, experiencing growth rates double those of the rest of the world, with online retail surging ahead at triple the growth rate[2]. However, retail has had its struggles in many Asian markets in recent years. This is largely due to the economic slowdown, declining brick-and-mortar sales, and the rise of e-commerce.

One might assume that brick-and-mortar stores are sliding toward obsolescence. However, established online players like Taobao, Zalora, and Bukalapak have invested in a physical presence, be it temporarily or for the long run. This suggests that there’s more than meets the eye. These players have recognised that the key to retail success is not pitting one shopping format against another—it is unifying them all. It is also where their omnichannel customers are buying, to maintain or grow market share.

For this reason, more is being done in Asia to refresh the retail experience. For example, the Singapore Government has been rolling out a Retail Industry Transformation Map (ITM). Envisioning a reinvigorated retail scene, the roadmap calls for helping local players integrate online and offline data, as well as leverage machine-learning algorithms to drive revenue and productivity. In Japan, Softbank introduced “Pepper”, a humanoid robot that can read human emotions and assist customers in Softbank’s mobile phone stores.

The promise of AI in retail is its ability to provide a shopper with a feeling of personal communication—or hyper-personalisation—across all shopping channels at scale, thereby increasing their engagement with the brand.

What's AI got to do with it?

In APAC, 79% of consumers are saying it is very important for companies to treat them as a person—and not just a number—to win their business[3]. Prices, promotions and advertising messages must now be automatically customised for each person, no matter how many customers a business has.

Retailers must also go beyond broad-based segmentation. Too often, in-app or website recommendations are based on static demographic or transactional data that miss the mark. For example, not all middle-aged males love golf, cars and fishing. Mobile devices, with their small screen sizes, also make personalisation even more critical. According to Criteo’s Q3 2018 Global Commerce Review, mobile devices now account for more than half of all online retail transactions in APAC.

Machine learning—a subset of AI—comes in handy to help retailers create detailed ‘profiles’ for each user and then tailor experiences that cater to distinct buying personas. Machine learning is also capable enough to deliver the same personalised, highly relevant experience to the individual, making it channel-agnostic. Its predictive nature is also what makes it a powerful tool for retailers, allowing them to use past and existing customer data to determine future trends and customer behaviours.

Research shows that top-performing companies are more than twice as likely to be using AI for marketing[1]. It therefore goes without saying that the best retailers must enable account-based marketing and meet customers on customers’ own terms—by showing them the products they want to see and serving the content they need to make informed decisions and astute purchases.

Early days for deep learning

Another subset of AI, deep learning, is designed to continually analyse data with a logic structure that mimics how a human would draw conclusions. For example, if a flashlight had a machine-learning model, it could be programmed to turn on when it recognises the audible cue of someone saying the word “dark”. With a deep-learning model, it could go further and intuitively figure out that it should turn on with the cues “I can’t see” or “the light switch isn’t working”.

Ensuring that a deep learning model does not derive and act on incorrect conclusions remains a tricky prospect. We must therefore better understand why these models work the way they do and their ability to generalise, and test their fragility, before committing to widespread applications in the retail space.

Alibaba, which recently announced a local research institute with Nanyang Technological University, is exploring the development and deployment of deep learning solutions across various sectors in Singapore, including retail.

My company is investing in deep learning research for advertising through our Criteo AI Lab in Paris, aiming to power new kinds of experiences across each touchpoint in the shopper journey, and eventually enhance sales.

Retailers are looking to dive deeper into the relationship between shoppers and the brands they interact with. AI adoption in retail is still gaining its foothold. However, it is important for retailers to leverage the technology to personalise the shopping experience as much as possible. Customer needs and preferences will continue to evolve, and AI will help retailers keep up with these changes more efficiently.

---

Alban Villani is managing director for SEA-Pacific at Criteo.

Sources