Why Omnichannel Retailers Adopt AI-Powered Search

By Marie Griffin

Against the headwinds of Brexit uncertainty, a soft economy, margin deflation and shifting consumer shopping habits, UK retailers must continue their omnichannel evolution in order to satisfy customers and stay competitive.

Internet sales represented 18% of UK retail revenues last year, with ecommerce sales up 14% over the prior year, according to a House of Commons briefing. In a study of UK shoppers from Periscope By McKinsey, 60% of respondents said that they shop in stores and online to a similar extent, while 21% shop mostly online and 4% shop online exclusively. Just 15% shop mostly or only in stores.

Omnichannel capabilities are essential as customers move effortlessly from one channel to another, expecting retailers to accommodate shopping preferences that fluctuate moment by moment. And retailers must respond to consumers who have become much more demanding.

Researchers from Salesforce and Publicis Sapient surveyed 6,000 adults in six countries, including the UK, and 67% of respondents said that their standards for good retail experiences are higher than ever. Consumers also want retailers to delight them continually with exciting new products. Whether they are visiting a website or store, 69% of shoppers surveyed said they expect to see new merchandise whenever they visit.

Meanwhile, the UK’s top online retailer, Amazon, is increasing its share of retail sales. According to GlobalData, 5% of all retail revenue in the country goes to Amazon, and half of British consumers purchase on Amazon every month. Having locked in the loyalty of 26% of the population with its Prime program, Amazon is set to expand its brick-and-mortar footprint in the UK. Already operating seven Whole Foods units, the company is reported to have secured space for its first Amazon Go shops in London.

The acceleration of AI development

Given the challenging business environment and shifting consumer habits, retailers are looking toward technology fortified with artificial intelligence (AI) and machine learning to boost supply chain efficiency, personalise customer relationships and fuel growth.

“Retailers, seeking any advantage they can in an increasingly competitive marketplace, are turning to AI technologies,” writes Matthew Bennett, partner, technology and media at CMS UK, in a report titled Disruption 2.0—here we go again.

As AI becomes a popular buzzword for technology vendors, retailers may find themselves confused by conflicting claims. There’s a reason for that. “There’s no single accepted definition of ‘artificial intelligence,’” according to  ExplAIned, a Guide for Executives, from Accenture Applied Intelligence. “And that’s because AI as we know it isn’t really a technology in its own right at all. In reality, it’s a collection of different technologies that can be brought together to enable machines to act with what appears to be human-like levels of intelligence.”

The idea that computers would eventually be able to “think” dates back decades, most famously to British mathematician Alan Turing, who developed the Turing Test whereby a computer would demonstrate its “sentience” by being indistinguishable from a human being when answering certain questions.

Within this decade, though, the development of AI has been accelerated by more powerful computing hardware, the availability of cheap storage in the cloud, improvements in mathematical formulas called algorithms and vast amounts of data for training those algorithms. As Ray Kurzweil, best-selling author, inventor, futurist and AI guru, put it: “There's a motto in the field that says life begins at a billion examples.”


Using AI to power search

Retailers must prioritise their investments as they implement AI projects, and search has become one of the most popular use cases. In a global survey of retailers and consumer goods companies, 40% of respondents that had adopted AI for at least one application were using it to provide relevant search results, according to Consumer Experience in the Retail Renaissance, a report by Deloitte Digital and Salesforce.

Capgemini Research Institute’s global survey of retailers implementing AI found a similar pattern. “Among the various categories of AI use cases for customer-facing functions, we found that personalised search, targeted recommendations and chat bots for answering customer queries were the most common implementations.”

AI and machine learning produce search results and recommendations in an entirely new way.

With a traditional rules-based search system, explicit programming directs the search. Think of it like a spreadsheet, where product categories have subcategories and descriptions that are linked together in a linear way. When a search does not follow the logic that has been programmed into the search engine, the results will be poor.

For example, when a shopper types “leopard print” into the search box, she might only see wall art with animals instead of apparel with a pattern similar to leopard fur. This is because the logic of the search separated the words leopard and print, and the word print was associated with the wall art and leopard was associated with animals.

Ultimately, the problem with rules-based systems is that it is impossible to anticipate all the different words and combinations of words a searcher might use, not to mention misspellings and other ordinary human mistakes that could cause a rules-based search to break down.

AI returns search results based on signals, which is data collected from every user interaction on the website. This includes queries, clicks, previous similar searches, product popularity, page views, add-to-carts and purchases. All of this data teaches the AI how to anticipate a user’s intention based on actual behaviour. AI is not limited by the ways humans understand language or concepts; it follows the data wherever it leads.

One example of the difference between rules-based and signals-based search comes from Peter Curran, president-cofounder of Cirrus10 and a consultant on the business and technology of ecommerce: “People will type the word ‘dysentery’ into the search box for a home goods website, even though it makes no sense for an intestinal infection to be associated with any of this retailer’s products,” he said. “But by going through enough sessions, the machine learning discovers that people who search for dysentery buy Dyson vacuum cleaners, and it starts automatically returning that result. A human merchandiser would never think to create a rule like ‘dysentery=Dyson,’ but the pairing occurs in actual searches for an ‘illogical’ reason. It’s because of the iPhone’s autocorrect feature.”

Dyson search

Users of the search box are more likely to buy

Retailers have traditionally viewed search as a tool that improves the customer experience — making it easier for the customer to find an item when they come to the site with a product in mind — but the real value of better search is that it produces better bottom-line results.

When a visitor uses the search box rather than browsing the site, that person is much more likely to become a buyer.  Conversion rates are on average 139% higher when search is used, as compared to the overall retail conversion rate, according to IMRG.

“Retailers invest millions in content management systems (CMS), ecommerce platforms and analytics solutions, but for some reason they underfund or don't consider search,” observes Grant Ingersoll, chief technology officer for Lucidworks. “But search gets the user from a query in the CMS to conversion or purchase in the ecommerce platform. This is where the money is made.”

Visitors who use the search box might be planning to buy a product online, but they also might be doing research in advance of a visit to a brick-and-mortar store. In any event, they will go to a competitor if they do not find the product when they use a retail brand’s website search — even when the product they want is available.

“Even if a purchase ultimately occurs offline, the role of online interactions remains critical,” according to The Era of Ecommerce: Capitalizing on the New Customer Journey from ClickZ and Catalyst, based on a survey of 750 consumers in North America. “We found that 74% of consumers research online and then purchase in-store either ‘always’ or ‘sometimes.’”

Consumers are more likely to start their shopping journey at a retail site when they know what they are looking for, according to the study. Half of those searches start on a retail site and half with a general search engine such as Google. In comparison, when shoppers are unsure of what they want, only 38% of journeys start on a retailer site and 62% start with a general search site.


Delaying AI adoption may be risky

There is risk involved in implementing any new technology, but retailers waiting to deploy AI are postponing exponential benefits. “Well-planned AI-driven solutions get more effective over time in a non-linear fashion,” according to the FitForCommerce Annual Report 2018. “They are self-leveraging. So, the ROI you realize two years from today will be vastly greater if you start tomorrow than if you start six months from now.”

Although AI is considered a buzzword in many circles, it’s still possible that organisations are undervaluing the potential impact of AI, asserts Andrew McAfee, principal research scientist at the Massachusetts Institute of Technology, in an interview with McKinsey Publishing. “Even though they see a lot of disruption coming, I still think that many really smart, well-managed companies are underestimating the scale, scope and speed of disruption this time around,” he said.

Marie Griffin, Lucidworks

Join thousands of other Online Retail professionals

Get unique insights straight to your inbox for free, and improve your understanding of online retail. Subscribe to Online Retail Weekly now.

Webinar Scroll Banner
Join thousands of other Online Retail professionals

Get unique insights straight to your inbox for free, and improve your understanding of online retail. Subscribe to Online Retail Weekly now.

Webinar Scroll Banner