On-site Search: Your Tool to 10X User Experience and Product Discovery
Once upon a time, on-site search was just a box in the corner of your homepage—something a user clicked when navigation failed, or when they already knew exactly what they wanted. Fast forward to today, and search has become one of the most important (and misunderstood) features of the eCommerce experience.
I recently sat down with Matt Eisnor, senior director global alliances at Algolia, to unpack what’s changed, what’s broken, and what’s next for on-site search. What emerged was a layered conversation that moved beyond algorithms into business strategy, user psychology, and the growing role of AI in product discovery.
From Keyword Match to Conversational Understanding
The first major shift we tackled was how search itself has evolved. Not just as a tool, but as a user behavior.
“In 2020, the average query was two words,” Matt shared. “Today, we’re seeing five-word searches become common, especially in complex catalogs.”
That growth isn’t random. It reflects larger behavioral shifts—from the rise of swipe texting and mobile-first habits to the normalization of talking to AI tools like ChatGPT. When users start treating on-site search more like a conversation and less like a command line, the complexity of interpreting intent increases exponentially.
Consider a query like: “organic fertilizer for a shade garden in Maine.” That’s seven words and multiple intent signals: organic (natural, not synthetic), fertilizer (product), shade (growing condition), and Maine (regional climate). A decade ago, search tools might have ignored half of that. Today, AI-powered search platforms are expected to parse and act on all of it.
Search as a Product Finder—and Beyond
We often think of search as a means to an end: I type something, I get what I want. But the truth is, on-site search plays multiple roles, and understanding those “jobs to be done” is crucial to evaluating and improving performance.
At Algolia, Matt described three key jobs for any modern search system:
Understand the query and user intent.
Retrieve results from the catalog, content systems, or external data.
Deliver results in a meaningful and timely way, across devices and formats.
But these three layers only scratch the surface. Search is also:
An education platform: it helps users understand what’s in your catalog and what’s not.
A discovery engine: surfacing complementary, adjacent, or alternative products.
A storytelling device: shaping how users perceive your brand’s breadth, organization, and value.
“Search is the easy part. Discovery—that’s the hard part.”
Matt Eisnor Senior Director, Global Alliances at Algolia
When Customers Outgrow Your On-site Search Experience
One of the most revealing moments in our conversation was discussing what happens when search fails. Not in a catastrophic sense—but subtly, silently, in ways that erode user trust.
There’s a quiet but telling metric: users turning to Google to search for products on your site.
“I’ve seen it in Search Console,” I told Matt. “Searches like ‘[retailer name] product name’—that’s a sign users couldn’t find it onsite. They’re using offsite tools to do your job.”
Matt agreed: “This was the number one use case in a recent RFP we responded to. The client said, ‘We don’t want our customers going back to Google to find products we know we have.’”
This fail state is more common than you might think—and it’s a signal that your onsite search isn’t just underperforming, it’s actively pushing people away.
The Power and Pitfalls of Personalization
As eCommerce expectations rise, so too does the need for intelligent personalization. But while the concept is easy to endorse, execution is tricky.
Done well, personalization feels natural, helpful, even invisible. Done poorly, it’s awkward at best—and alienating at worst.
We laughed (and groaned) about Amazon’s infamous recommendation engine, which often spams users with ads for products they’ve already bought. “It’s like shouting, ‘We know what you bought! But we didn’t think about what that means!’” I said.
The smarter version of personalization understands what comes next. If I buy a tent, maybe I need sleeping bags. If I order a printer, show me ink—not more printers.
Matt explained how Algolia enables this kind of depth by letting merchants weigh various signals—margin, inventory availability, geographic trends, popularity—and combine them in real time. “It’s not just about relevance,” he said. “It’s about the right relevance, for both customer and business.”
Why AI Still Needs a Human Touch
While AI has supercharged what search can do, it’s not a magic bullet. And the more powerful it gets, the more oversight it requires.
“A lot of AI tools are overly confident,” Matt said. “They hallucinate. They return results you can’t always explain. That’s dangerous when the goal is trust.”
That’s why transparency and control matter. Algolia offers full visibility into why search results were ranked the way they were, and gives merchants the ability to adjust and reweight them. This is especially critical in B2B and manufacturing, where industry jargon, part numbers, and acronyms need domain-specific logic.
If your AI tool can’t understand the difference between “PST” as a voltage class and “PST” as a time zone, you’ve got a problem.
Implementation Is Just the Beginning
Too often, teams treat search as a set-it-and-forget-it solution. They implement a new tool, see immediate improvements, and move on to the next fire.
But the biggest value in platforms like Algolia doesn’t come from flipping the switch—it comes from iterating, analyzing, and improving month after month.
Data connections between marketing, inventory, and customer behavior
We don’t just fix the search experience. We use search data as a feedback loop to improve the whole site.
What’s Next: Agentic Commerce and AI Orchestration
As we looked ahead, Matt introduced the concept of agentic commerce—a phrase you’re likely to hear more often in the next couple of years.
“It’s about orchestrating multiple AI tools into a single experience,” he said. “It’s already starting to happen.”
That could mean a future where search, recommendations, chatbots, and even support are all AI-powered and working together. But even as the technology evolves, our core responsibility stays the same: listen to the customer, understand their intent, and help them make a confident, informed decision.
The Hard Part Is Human
At the end of our conversation, Matt said something that stuck with me: “The technology part is the easy part. Your part—the UX strategy, the content, the customer understanding—that’s the hard part.”
And it’s true.
Technology can enable, but it doesn’t replace the need to ask real questions, observe real behaviors, and adapt thoughtfully. That’s what makes search a strategic engine—not just a box on the page.
If you’re ready to treat search as more than a utility, we’re here to help.
Let’s turn your search bar into a business accelerator.
Dane Dickerson
Dane Dickerson is Human Element’s digital marketing team lead. His 11+ years in marketing include adventures in SEO, online advertising, email marketing, UI/UX, and remarketing. Every time he brings up the AOL / Time Warner merger or rants about trademark law he has to put a nickel in a jar. He resides in Central Arkansas and moonlights as an event photographer and live audio technician.
