Search Relevance is Broken: Why Product Discovery Software Made for E-Commerce is Better

Relevance has historically underpinned much of e-commerce search and product discovery strategy. The saying goes, “Reach the right people in the right place at the right time,” and relevance is typically believed to be at the foundation of that objective, guiding retailers and brands as they aim to surface the right product in the right context to the ideal target audience.

The centrality of relevance to digital marketing and e-commerce is why it’s so urgent for the industry to reckon with an important realization: the concept of relevance is broken when it comes to e-commerce.

As many technology providers use it, relevance stands for little more than semantic matching. A consumer searches for a shirt or shows interest in shirts, and an algorithm surfaces a product that matches that description.

Relevance as it’s operationalized by e-commerce and marketing teams should be much more impactful than that. It should optimize for surfacing products that are not just semantically relevant to a consumer's query, but attractive to where the shopper is in their journey and the seller’s goals, such as revenue, profit or conversion optimization.

Let’s dig into why the conventional conception of relevance is failing e-commerce shoppers and sellers, the digital tools needed to support a more comprehensive and nuanced approach called attractiveness, and how this strategic reframing stands to benefit sellers and marketers in the long run.

Why We Need to Rethink E-Commerce Relevance

The main problem with the concept of relevance in e-commerce today? It’s defined by the standards of content search, not product discovery. But content search and product discovery are two very different digital journeys, sparked by different motivations and defined by different satisfactory outcomes.

While relevance based on semantic matching makes sense for finding information online, it's not directly translatable to e-commerce product discovery and the buying journey.

When looking for products, consumers want results that are not just relevant but attractive — i.e., aligned with their desire to make a purchase. Defining relevance by content search standards leads to product results that are semantically relevant to a search term or category page, but often not what the consumer wants to buy. Consider, for example, a search for “chips” on a large general retailer’s site, which returns over a thousand results of very different types: tortilla chips, potato chips, chocolate chips, wood chips, computer chips, poker chips.

Even queries that seem to have less ambiguity in intent can still give consumers the idea you don’t carry what they want to buy. They can feature last season’s style and no longer be popular, or they may be overly expensive, or just generally not something most customers like. Or they can just be the wrong products for a particular shopper.

Imagine a health-inclined buyer going to the “chocolates” category and seeing only cheaper products loaded with preservatives and sugar on the first page. Or a shopper who wears size small searching for “shirts” and finding that every result they click on is only stocked in size medium and above. The results are perfectly relevant. They match the query. But they aren’t the products those users want to buy. This is often worse than showing irrelevant products because those can at least lead to the shopper trying a similar query. The relevant but unattractive product tells the shopper that you’re showing them the products you carry that are relevant to their query. They just aren’t products the shopper wants to buy and so the shopper should try searching at a different site.

What’s the solution? Brands and retailers need to move beyond relevance. The goal isn’t to show the shopper something technically relevant. It’s to show them something they will consider buying.

E-commerce product search requires a more nuanced understanding of the shopper’s product discovery journey and an awareness of the specifics of what intent looks like for shoppers at different stages of that journey. We need to replace relevance with attractiveness, which means taking into account everything you know about the product catalog and the journey of a particular shopper, and then showing the products most likely to meet at the intersection of what the shopper most wants to buy and what the store most wants to sell.

How Technology Must Change to Solve for Attractiveness

Returning attractive results is no easy task. Consumers are overwhelmed with choice when it comes to what and where to buy online. If a site search experience fails to quickly surface products shoppers want to buy, nearly half of consumers say they’ll go elsewhere to make a purchase.

How can sellers ensure that their product discovery technology can process millions of queries a day, while still providing results that are attractive to individuals at different points in the buying journey?

Machine learning capabilities and artificial intelligence-first technology is the key. An AI-powered product discovery platform, which learns from human interactions and can be tailored to optimize for the intersection of consumer interests and the business metrics brands and retailers want to solve for, is equipped to understand and address the nuances of consumer intent in a digital context.

Machine-learning technology allows a search and product discovery engine to provide results optimized for attractiveness. Drawing on data from user behavior, AI puts together a picture of individual and collective consumer interests and orders results accordingly.

For example, if a consumer indicates they prefer Apple products and go to the laptops category page, they’ll be more likely to see MacBooks. If they typically buy dog food at an online pet store and then click on a pet food sale page, they’ll be more likely to see dog food than cat food. If they indicate they’re based in Florida and search “pants,” they won’t see the same jeans made of thick denim that a user in New York might see during the winter. The results they see will still look relevant, but they’ll also be attractive and more likely to lead to an actual purchase.

Older, text-based matching models that power content search aren’t equipped to respond to the particular challenges of parsing consumer intent — and neither are retrofitted “solutions” that try to patch over the shortcomings of the content-centric search model. Delivering attractiveness in e-commerce demands technology built specifically for e-commerce and based on AI.

What E-Commerce Sellers and Marketers Gain From Leveraging Attractiveness

Attractiveness cultivates the kind of product discovery experience that supports sellers’ short- and long-term business goals.

An e-commerce product discovery engine that can anticipate and prioritize customer needs transforms the buying journey into a smooth, holistic experience. Adding nuance and personalized context to search results and recommendations cultivates the delight that drives conversion and customer retention. According to a 2021 survey, 78 percent of consumers were likely to return to make future purchases from companies that offered personalization.

Using data to consider attractiveness at multiple points in the customer journey allows brands and retailers to more effectively shape merchandising strategy in the long term, optimizing ranking according to the key performance indicators that matter to them.

With AI driving search, less time needs to be spent on manually executing on merchandising strategy. Less time spent on rote, manual tasks like identifying synonyms or boosting and burying items means that merchandising teams are better equipped to not just drive conversions but minimize returns, balance inventory, and contribute creatively and strategically to the business.

Supporting a higher standard than relevance in e-commerce product discovery and search is a boon to stakeholders on all sides of the e-commerce experience. Advances in machine learning and AI will continue to revolutionize the e-commerce landscape, but they’ll always only be as good as the metrics they solve for. Don’t just solve for relevance. Solve for attractiveness and see your revenue and customer satisfaction soar.

Eli Finkelshteyn is the co-founder and CEO of Constructor, an AI-first e-commerce search engine.