How Retail Changes When Algorithms Curate Everything We Buy

By Matt Dallisson, 07/01/2019

Some established retailers already offer services to help customers find the most suitable products among those they supply. Amazon collects user reviews and makes customized suggestions based on learning algorithms. In the United Kingdom, womenswear retailer Topshop and department store John Lewis partner with online engine provider Dressipi to create personalized outfit recommendations based on initial profiling followed by machine learning applied to preferences.

A new generation of retail choice engines will work more clearly on behalf of customers by offering transparency, neutrality, and an unlimited catalog. Just as flight intermediaries such as Google Flights, Hopper and Skyscanner find the lowest possible prices, agnostic digital retail curators could direct consumers to the retailer offering the best deals — or advise them to delay a purchase when a promotion is likely. In the same way that Expedia makes bookings directly with hotel chains, these digital curators could negotiate terms directly with manufacturers.

When this happens, we expect that retail curators will become an industry on their own, changing the structure of the retail sector and capturing a significant share of retail sales. Consider this: Our research shows that the three biggest online travel intermediaries — Expedia, Booking Holdings, and C-Trip — accounted for nearly 20 percent of the global travel market’s $1.3 trillion in sales in 2017. We estimate from company reports that they took between 3 and 20 percent of this share as fees, amounting to over $25 billion in net revenues. They forecast annual growth rates of between 20 and 45 percent.

We believe that retail curation could follow a similar trajectory. To be sure, technology advances and differences between travel and retail curators make direct comparisons difficult. But digital retailers like Amazon and Alibaba are already blurring traditional retail boundaries and consumers have demonstrated a large and growing appetite for digital retail. It seems fair to assume that retail curation will ultimately look more like travel curation than not.

Below, we explore three types of digital curating engines that are emerging.

Market Mappers

Market mappers organize a vast range of choices in a compelling, transparent way in the travel industry. Companies such as Skyscanner and Booking Holding’s Kayak optimize thousands of route and carrier combinations to give users cost- and time-effective options from a complex set of variably priced airline tickets.

The same principle works in retail — but on a far smaller scale up to now. Google Shopping, for example, will search for the best deal on a coat of a certain design, color, and size. It finds TV screens and dishwashers within a certain price range. And it shows options to buy related items like detergent. Google Shopping then sends the customer to the appropriate online store.

That makes Google Shopping useful for a single, high-priced item, such as the dishwasher or TV screen. But in the future, we expect these agnostic curating engines will scour the planet to create baskets with optimal combinations of low-cost household goods – much as Skyscanner and Kayak now find the best options for trips that require multiple flights. These engines will consider not just products’ price but delivery and other potential costs to suggest the best deals.

Once that happens, market mappers will be able to find the best combination of items for a full online shopping trip, where the customer pays the shopping service for a variety of goods in one go.

Digital Personal Shoppers

A step beyond market mapping is to tailor offers to individual shoppers. Currently, a number of wardrobe management services— including Thread, Dressipi, Trunk Club, Style Lyrical, and Cladwell — do this to a certain extent using a combination of artificial intelligence and human interaction.

Thread, a clothes shopping service so far just for men, for example, populates different style templates with clothes from an array of brands. Customers indicate which templates they like and provide information about themselves, including the kind of clothes they already have, their desired price range, and whether, for instance, they want a daring or a conventional look.

A human stylist uses these to design a basic look for a customer, complete with colors. Then an algorithm will pick out a pair of pants in the customer’s size from a specific brand, taking the customer’s preferences into account, such as whether he has previously rejected jeans or indicated that he likes chinos. Each stylist can thus look after tens of thousands of clients.

The key technology is learning algorithms, which acquire knowledge about a customer’s stated preferences and purchase behavior in order to predict what she will want and develop tailored recommendations. Pioneered by Netflix, among others, these types of algorithms get better at their predictions as the customer approves or ignores successive offers.  Ultimately, digital human-machine services like Thread could end up playing a key role that department stores currently do, curating a vast range of clothing or other consumer goods to make it easier for customers to find just what they want.

Review Aggregators

Today, most e-commerce retailers integrate user reviews into their sites. Amazon and Walmart have huge banks of these, while smaller retailers can access a wider set of reviews than just those of their immediate customer through aggregators such as Bazaarvoice. Makeup Alley gathers make-up reviews and tips from sources such as blogs.

While Amazon’s vast range of reviews provides it with a strategic advantage, it has had to overcome challenges related to the trustworthiness of reviews. To increase faith in customer feedback, two years ago Amazon ceased allowing reviews posted by anyone who had in exchange received a free product sample directly from the supplier. Instead, suppliers must now distribute samples through Amazon’s Vine program, which acts as a middleman between reviewers and vendors and allows freebies to pass from one to another so long as there is no direct contact.

As consumers buy more types of product online, the importance of fair reviews across platforms will increase. We expect this will lead to the rise of review aggregators across platforms in a broadening range of retail segments. As Yelp and TripAdvisor do in travel, these will filter reviews so that customers read those by people with similar purchase goals, which are most likely to be relevant and trustworthy.

Curating Retail

Of course, developing these engines for retail will probably be harder than it has been for travel. Flights are already displayed in a standardized way on airline websites, making it relatively straightforward to trawl offers, make comparisons, and assemble packages. In food and apparel, however, prices and specifications may or may not be available depending on the retailer, the brand, and the type of product. Moreover, flights are relatively expensive, so it’s been worth it to the travel platforms to help a customer choose just one. For other product and services classes the success of digital curators will depend on higher volumes given the lower revenues per item.

But retailers are already beginning to pioneer shopping models that show these types of curating engines are attractive to shoppers. And when retail curators gain momentum, the model could scale well globally, particularly in segments where products tend to have the same specifications in all markets, such as technology and some kinds of apparel.

Once that happens, retailers, just like travel agents before them, risk becoming back-end utilities serving new intermediaries. They might supply a ready catalog of standard products and struggle to differentiate their offer. For some products, they could even be cut out altogether. Instead, agnostic digital “food butlers” might advise people on dinner choices ranging from a recipe cooked from scratch, to a frozen premium ready meal, to a takeaway from a nearby restaurant. Online digital personal shoppers could suggest mixes and matches of inexpensive accessories from fast-fashion outlets with garments from higher-end boutiques that are built to last and will optimize the cost per use over several seasons.

The best way for retailers to protect themselves: Give customers good reasons to keep coming to them directly.