Search: Intent, Not Inventory

When I enter a physical store, my intent is “I want to buy X” or “I’m looking for something in order to do X.” It’s not “Show me the products that satisfy characteristics X, Y, and Z.” My intent is to find what I need, not to learn what products the store sells or how the store organizes them.

And yet, when I use a search engine to shop online, I’m forced into an inventory-centric view. The search engine interprets my keywords as a request to show me the products containing those keywords — or, if the search engine understands my query, the product attributes denoted by those keywords. When I use a site’s browsing and faceting functionality, I find myself filtering its inventory based on the structured data used to organize the catalog. Either way, it’s an inventory-centric experience.

That’s not how search should work, especially on a shopping site. The search experience should center on the searcher’s intent, not the inventory.

Think Outside the (Search) Box

To be fair, a search for “shoes” serves just as well as shorthand for “I want to buy shoes” (intent-centric) as it does for “Show me shoes” (inventory-centric). This shorthand is convenient: despite the excitement around natural language interfaces, most people prefer to type a couple of keywords than to type a long sentence. Perhaps searcher preferences will evolve with the emergence of voice search, but we’re not quite there yet — at least for shopping.

Regardless, the search experience shouldn’t just be typing a few words into a search box and then slogging through a list of results. An effective and delightful search experience is one that progressively elicits and elaborates the searcher’s intent, like a helpful assistant in a physical store.

Crafting the Search Journey

What makes a search experience intent-centric? Here are a few methods that well-designed search experiences use to help ensure that the search journey centers on the searcher’s intent rather than the inventory:

• Autocomplete anticipates the searcher’s needs.
  Many search engines treat autocomplete as an afterthought, populating suggestions with the most frequent queries as shortcuts for searchers. But autocomplete should be more than just a convenience. It should also help searchers express their intents, guiding them to successful search experiences. Autocomplete suggestions should anticipate likely searcher intents and present query suggestions that searchers can easily recognize as expressing those intents. As a corollary, autocomplete suggestions should be unambiguous, specific queries that lead to successful search experiences — something a search engine can learn from behavioral data.
• The search engine understands the searcher’s vocabulary.
  The catalog’s vocabulary is the basis for the search index and determines how the search engine responds to queries. But the catalog’s vocabulary doesn’t always align with searchers’ vocabulary. It’s important that search engines deliver useful results when searchers use their own vocabulary, regardless of the language used in the catalog. That means using synonyms and other query rewriting methods to translate the query from the searcher’s language to that of the search index. Indeed, even if the catalog does not contain any inventory matching the searcher’s intent, the search engine should recognize this fact and respond appropriately, e.g., “Sorry, we don’t have any X, but we do have these related products.” Query understanding starts with understanding the searcher’s vocabulary.
• The search experience centers on searcher demand, not supply.
  Many search engine features are supply-driven. In particular, counts are supply-driven by definition, whether they are the number of results matching a search query or the counts associate with facets. Supply, however, is inventory-centric rather than intent-centric. An intent-centric search experience should center on demand. For example, the top facets for search query should be those associated with the highest demand, not the highest supply. Similarly, determining whether a query is broad or ambiguous should be based on demand entropy, rather than supply entropy. Building a demand-driven experience requires learning from user behavior, rather than relying solely on the search index.

These suggestions offer a taste of how to center on searcher intent rather than inventory. Crafting an effective, delightful search experience requires an intent-centric mindset driving every aspect of the design process.

If you’re building or managing a search engine, remember that searchers don’t care what you have in your catalog or how you organize your inventory. They care about what they want. Make it easy for searchers to express their intent, and craft the search experience around their intent. That way, they’re more likely to find what they want — and to buy it.