In one of U2’s best-known songs, Bono sings dolefully, “I still haven’t found what I’m looking for.” He’s not alone. On many retail websites, shoppers longing to find an item with ease and precision also end up feeling unfulfilled.

In one survey of U.S. shoppers, a solid majority (60%) reported being frustrated with irrelevant search results. In another, almost half (47%) of online shoppers complained that “it takes too long” to find what they want, while 41% have difficulty finding “the exact product” they are looking for.

Consumers’ frustration with poor shopping site search is no small matter. The fact that shoppers can search independently for products online, as opposed to waiting for a retailer to present them, is at the heart of the ongoing transformation of retailing.

60% of shoppers reported being frustrated with irrelevant search results

Recommendations Create the Shopper Journey

Analysts at Deloitte Consulting identified the “trend for consumers to take their own lead in the shopping journey” in a report called The New Digital Divide. “A significant number of consumers want to manage the journey themselves, directing the ways and times in which they engage retailers rather than following a path prescribed by retailers or marketers.”

Precise, personalized, and speedy search—and the companion function of product recommendations—is becoming a key differentiator in ecommerce. Even Amazon, where 47% of shoppers already start their product searches, is working on the next-generation search experience.

Amazon’s new service called Scout is based on products’ visual attributes, reports CNBC. “It is perfect for shoppers who face two common dilemmas: ‘I don’t know what I want, but I’ll know it when I see it,’ and ‘I know what I want, but I don’t know what it’s called,’” said a statement Amazon sent to CNBC.

As leading-edge ecommerce sites like Amazon train consumers to expect better search, shoppers will have even less tolerance for a mediocre search experience.

‘What’s New?’ Is Still the Leading Question

Once customers have made a purchase, retailers have the knowledge to use to entice them back. A simple and effective way to do that is with automation; Amazon’s Subscribe & Save program, for example, provides a discount to shoppers who sign up for regularly-scheduled automatic delivery of certain items.

However, because consumers overwhelmingly want to see what’s new when they interact with a retailer, site search and personalized recommendations provide ecommerce with the greatest opportunity to capture new shoppers or to introduce existing customers to additional products or categories.

In fact, 69% of consumers responding to a Salesforce and Publicis.Sapient survey reported that it is “important” or “very important” to see new merchandise every time they visit a physical store or shopping site, and three-quarters of shoppers are using new site search queries online each month.

This explains why more than half (59%) of the top 5% of best-selling products on e-commerce sites change every month, according to the report. “That means retailers and brands can’t sleep on analyzing shopper searches and delivering the ever-changing items they seek in real time.”

Machine Learning: Know Your Customer

E-retailers are increasingly using Artificial Intelligence (AI), specifically Machine Learning (ML) and Natural Language Processing (NLP), to help shoppers discover what they want, perhaps before they know themselves.

Engagement pays off: 6% of e-commerce visits that include engagement with AI-powered recommendations drive 37% of revenue

AI-driven personalized recommendations can also provide a big payoff for retailers. A survey by Salesforce and Publicis.Sapient found that “6% of e-commerce visits that include engagement with AI-powered recommendations [drive] an outsized 37% of revenue.”

“The best way to understand your customers’ needs is to actually track and listen to your consumer,” said Lasya Marla, Director of Product Management at Lucidworks. “You do this by tracking customer signals, what they click on, what they ignore, what they call things. Recording and analyzing signals is crucial to learning your customers’ likes and dislikes and their intent.”

Merchandising Expertise Still Key

While machine learning can automatically suggest products and help customers discover items they wouldn’t have otherwise, many brands, particularly lifestyle brands, are loathe to risk merchandising with machines while they have experts on hand.

According to Peter Curran, president of Cirrus10, an onsite search system integrator in Seattle, “we work with brands that want ML to eliminate the drudgery of search curation—synonyms, boosts, redirects, and keywording—but who still want to finesse the customer experience.

“The dance between brand and brand aficionado is filled with nuance that merchandisers tend to notice—and IT departments tend to miss. The role of the merchandiser is ready for transformation,” Curran continued. “Feature selection, entity extraction, embeddings, and similar concepts are currently the job of the data scientist, but that work can’t be done well without the cooperation of the business user. We need tools that allow business users and data scientists to cooperate on improved models and always allow business users to override automation.”

Next Steps for Brands and Retailers

For ecommerce retailers and brands thinking about upgrading and modernizing their search functionality, it’s critical to develop a strategy that is integrated with the organization’s long-term goals. There are many aspects to consider during the selection of new technology for site search, but these questions can help in the process:

  • How can we develop better algorithms and techniques to match keywords to products?
  • What do we need to automatically fix search keywords based on misspellings, word order, synonyms, and other types of common mismatches?
  • How can we enable our marketing and merchandising people to take control of search so that it supports the business, including promotions, inventory, and seasons?
  • What tools do we need to analyze search trends at both an individual and macro level so that we can adjust in real time?
  • What are the signals customers are sending—and how can we best capture them?

While it is true that looking for a new pair of boots or a specialized metalworking tool does not rank up there with the search for a soulmate that U2’s Bono sings about, the desire to find a certain item in an online store has an emotional component that is intimately connected to the shopper’s perception of that brand.

Ironically, technology can produce search results and recommendations that are so personalized that they enhance this emotional connection, giving the consumer the sense that the brand “knows me.” The choice of a platform for site search, then, will make visitors fall more deeply in love with a retail brand—or send them elsewhere to find what they are looking for.

Marie Griffin is a New Jersey-based writer who covers retail for numerous B2B magazines.