What is product discovery in ecommerce?

Ecommerce product discovery is how shoppers find the right product on a brand’s site, whether they browse categories, search, follow recommendations, or ask questions in a conversation. It covers shoppers who arrive knowing the exact product they want and shoppers who only know the problem they’re trying to solve.

In product management, “product discovery” means deciding what to build. This entry covers the shopping meaning.

Product discovery in one sentence

Product discovery is how a shopper gets from a need to the right product, whether they browse, search, follow a recommendation, or ask.

How shoppers discover products

Every store offers some mix of four discovery methods, and each one asks something different of the shopper. Browsing needs them to know the category, search needs the right words, recommendations need the store to know something about them, and conversation needs them to describe what they’re after.

Most discovery work on ecommerce teams goes into the first three rows. If search is your weak spot, start with our site search best practices. If shoppers can find the category but drown in results, read up on faceted search. Conversation is the newest of the four, and guided selling covers how stores put it to work.

Product discovery vs. site search vs. recommendations

Teams often treat these terms as one thing, but each covers a different part of the path to purchase.

Site search answers a query the shopper already has. It’s reactive, and it depends on the shopper typing something the catalog understands.

Recommendations predict what a shopper might want based on behavior: what they viewed, what people like them bought, what usually goes in the cart together. Website personalization extends the same idea to the whole page, changing banners, sort order, and content by visitor.

Product discovery is the full journey from “I’m not sure what I need” to “this one.” Search and recommendations are two of the tools that get a shopper there, alongside navigation and conversation. A store can have excellent search and still have poor discovery if the shoppers who arrive with a problem, and not a product name, have nowhere to go.

Why product discovery matters

Shoppers can only buy what they find, and on most sites a lot of them don’t find it. Baymard Institute’s benchmark of more than 170 ecommerce sites found that 56% of sites fail to adequately support users’ search needs, and about half of Baymard’s test participants prefer search over the main navigation when they’re hunting for a product.

The failures cluster around a particular kind of query. In the same research, sites had issues with use-case searches like “party dress” 43% of the time, compatibility searches like “Dell laptop adapter” 44% of the time, and feature searches like “volumizing paraben-free shampoo” 39% of the time. Those are the searches shoppers run when they know their problem but not the product name.

Searches that dead-end make it worse. Baymard also found that nearly 50% of sites give users no effective way to recover from a search with no results. A shopper who hits a blank page and no next step usually doesn’t try a second phrasing. They leave, and they take the purchase to a site or an AI tool that answered them.

A brand with 40 products can get by on good category pages. A brand with 4,000 SKUs, several fits, and a dozen compatibility rules needs every method in the table above working together.

Signs your product discovery is broken

You don’t need a new analytics stack to spot weak discovery. Most of the evidence is already sitting in reports your team owns.

High search exits. If a large share of shoppers leave right after viewing search results, the results aren’t matching what they asked for.

A climbing rate of searches with no results. Pull the actual terms. They’ll show you the words shoppers use that your product data doesn’t, plus the products they expected you to carry.

Chat full of “do you have…?” questions. When a shopper opens chat to ask whether a jacket comes in tall sizes, which sunglasses fit a narrow face, or whether a part works with last year’s model, they’ve usually tried to find the answer on the site first. Each of those conversations points to a discovery gap.

Returns driven by fit or compatibility. A return coded “didn’t fit” or “not compatible” is often a discovery problem that showed up late. The shopper found a product, just not the right one.

Shoppers who browse many products and buy none. Long sessions with lots of product views and no cart add can mean a shopper is comparing on their own because nobody helped them narrow it down.

How product discovery has evolved

Each discovery method was added to catch shoppers the earlier ones missed, and most stores now run several at once.

Category navigation. Early online stores borrowed the department-store model and arranged products by type in menus, so shoppers browsed the way they’d walk the aisles.

Keyword search. As catalogs grew, a search box let shoppers skip the menus, but only when they typed words that matched the product data.

Faceted filters. Filters for size, color, price, and other attributes let shoppers narrow a long list of results without starting a new search.

Recommendation engines. Collaborative filtering, which Amazon described in a 2003 paper on its item-to-item approach, made “customers who bought this also bought” a standard feature.

Semantic search. Search engines began using natural language processing to match the meaning of a query, so the right products can show up even when the query’s words don’t appear in any product title.

Conversation and AI agents. Shoppers can now describe a need to an AI assistant on a brand’s site, or to a general AI tool like ChatGPT, and get back a short list of products.

AI and product discovery

AI now shows up in every discovery method. Site search engines use language models to understand intent, so “dress for a beach wedding” returns dresses and not a list of products that happen to contain the word “beach.” Recommendation engines use AI to rank products by likely interest. On-site assistants hold the conversations described above.

Discovery is also moving off your site. In July 2026, 42% of US consumers said they’d used AI assistants for online shopping, according to Adobe, so some shoppers now meet your products in ChatGPT, Gemini, or Perplexity before they ever reach your site. Our guide to AI product discovery for ecommerce maps the four places AI now shapes discovery and where to start.

That gives ecommerce teams two jobs. Your product data has to be clean enough for outside AI agents to read, which our guide to the product feed for AI agents covers. And when those shoppers do land on your site, the conversation they started elsewhere should be able to continue there. Our pieces on agentic commerce and AI shopping agents go deeper on how agents change the funnel.

Metrics to track

These are the discovery metrics that tell you the most about where shoppers get stuck.

Search exit rate. The share of searches that end with the shopper leaving the site. Break it out by query type to see which kinds of searches fail most.

Rate of searches with no results. Track it weekly and review the terms behind it.

Search-to-cart rate. How often a search session leads to an add to cart. It’s a better read on search quality than clicks alone.

Assisted conversion rate. The share of orders where the shopper interacted with a discovery tool first: search, a recommendation widget, or a conversation.

Revenue per conversation. Total revenue attributed to conversations, divided by the number of conversations. It tells you whether chat is selling or just answering. Watch it alongside deflection rate, which measures how many questions get resolved without a team member. A high deflection rate with low revenue per conversation usually means the AI is answering questions but not helping anyone choose.

Share of conversations that happen before purchase. If this number is growing, shoppers are bringing discovery questions to chat, and the team or AI answering chat needs product knowledge to match.

Tie all of these back to your overall ecommerce conversion rate, since that’s where better discovery should show up.

Related terms

Guided selling. Asking shoppers about their needs and recommending the products that fit, through a quiz, a product finder, a human expert, or an AI assistant.

Faceted search. Search results that shoppers can narrow with filters built from product attributes, like size, material, or price.

Zero-results page. What a shopper sees when a search returns nothing. A good one suggests related products, categories, or spelling fixes so the search doesn’t end there.

Product information management (PIM). Software that stores and standardizes product data, so search, filters, product feeds, and AI assistants all work from the same attributes.

Merchandising rules. The settings that decide which products appear first in categories and search results, such as boosting bestsellers, new arrivals, or in-stock items.

Agentic commerce. AI agents that research, compare, and buy on a shopper’s behalf, within limits the shopper sets.

Conversion rate optimization. The practice of increasing the share of visitors who complete a purchase or other goal. Better discovery is one way to do it.