How AI Is Quietly Transforming E-Commerce Search And Discovery Gaylord Contreras | usagoldmines.com

In the age of Google and now ChatGPT, we take quality search for granted. We’re used to having information at our fingertips, but this ecosystem of instant gratification hides complex processes that enable us to find what we’re looking for.

Maybe because of search’s ubiquity, good search and discovery is often overlooked in e-commerce. But getting this element right can have a huge impact on your business. Effectively matching customers and products increases conversions and reinforces trust and loyalty as customers realize you’re relevant to their needs. Increasingly, AI is driving a revolution in search and discovery, being implemented by everyone from Amazon to Shopify to personalize search results and allow for deeper categorization of products that reaches beyond standard keywords.

Think of search as a two-way street, an interaction between your products and your customers. AI allows us to understand both points on this line better, and draw the shortest possible distance between the two. Here’s how AI can help update your search and discovery.

Moving Beyond Keywords

Search and discovery has traditionally been driven by keywords, returning products that are the closest match to a user’s chosen search words. The problems with this approach are obvious: firstly, it requires both accurate classification of products. It also requires your customers to know pretty much exactly what they want to buy! While keyword search remains at the heart of most algorithms, including Google and Amazon, in e-commerce it has contributed to a failure rate in first searches of up to 17% and over two-thirds of consumers seeing irrelevant results.

For an online fashion brand or food delivery app, keyword search is a functional starting point, as customers often begin searching with a decent idea of what they’re looking for, be it jeans or Chinese food. Brands must be alert to the diversity of customer needs, though, and keywords should include broader categories and common phrases such as ‘healthy food’, ‘quick delivery’, or ‘relaxed fit’.

At Atom.com, our focus is on brandable domains. These often contain unique naming styles that make discovery less linear than that provided by a simple keyword approach. The main selling point of our domains may be the metaphor or emotional invocations, so the connection to keywords is tangential or opaque. Across e-commerce, brands may have product categories where keyword relevance is unclear; moreover, an intelligent search algorithm should read between the lines of consumer keywords to present supplementary products that the customer may want without yet realizing.

AI and the Future of Search and Discovery

In September 2024, after months of testing, we fully integrated AI with our search. This enabled it to continuously critique and improve its own results. By training AI models to think like buyers, we were able to identify irrelevant results across thousands of keywords, helping us eliminate results that didn’t align with buyer expectations. Since this implementation, we’ve already had some clear improvements: 17.4% better engagement and 14.6% increase in conversions. Here’s how AI has enabled an optimized search and discovery process.

Deep Classification

AI enables a deeper and more complex keyword classification than could ever be afforded by a manual process. Our buyers typically start by searching broad industry themes, such as fashion or beauty, as they work on building brands within these categories. With tens of thousands of domain names in these categories, reliably recommending the most relevant names is impossible, even with a huge base of root keywords.

We’ve implemented AI to build not only a deeper set of interrelated relevant keywords, including themes, emotion and name style, but also to generate accurate use cases for our domains. This means that if a buyer searches for use cases such as ‘sustainable clothing brand’ or ‘natural skincare brand’, our search will now prioritize the discovery of names like ‘PurityCompass’ which are highly relevant to the brand intentions without being rooted in direct keywords. Instead of returning names with the user’s search terms forced in, our AI-based discovery process offers an intuitive solution based on end-use.

Deeper classification is essential for any e-commerce brand, even when many of your products may be found by more direct keyword search and it enables your search algorithm to return highly relevant answers even when buyer keywords are lacking specificity. Moreover, by offering customers diverse but relevant products, you will learn more about their behavior enabling individual personalization and accurate segmentation.

Connection With The Buyer Journey

Deep classification – and the accurate categorization it requires – is just one side of the coin of better discovery. The other side is your customers, those doing the searching. You must pair deep classification with data on buyer types and behavior. In other words, how buyers are browsing your marketplace. This will allow you to create meaningful associations based on real end-user search terms.

In most industries it’s likely buyers will return to use your search function multiple times before making a purchase, as well as after an initial conversion. Data gathered through these touchpoints can be built into detailed user profiles. For example, if a buyer searches for a certain keyword and shortlists 5 different names, our AI begins building dynamic associations based on search intent and those domain attributes. With a dataset of 3 million unique monthly visitors, we’re able to aggregate insights to improve the discovery experience across the board.

Deep classification is integral to this, as it allows our search algorithm to cross-reference the products that pique the buyer’s interest and present personalized suggestions that are based upon how similar users interacted with those names in the past.

Of course, every customer is different. Advanced machine learning allows our dynamic user profiles to adjust in real-time, based on search behavior. The result is ever-improving search results that hone buyers in on the right products at the right time.

Search Is About Data And People

Effective search and discovery requires a two-pronged approach. Firstly, a deeper, more sophisticated product classification. Second, understanding your customers and their search intent.

Once you’ve perfected both, search is no longer a coin toss for your customers but an accurate and effective tool for bringing the right products to the right buyers. Search should be able to handle customers at all stages of the buyer journey, and provide relevant and desirable results for those with different degrees of knowledge and information about your inventory. It can be the final step to finding and buying a product, or a mid-funnel test of just what you have to offer, so it truly is a core part of your business as an e-commerce provider!