Amazon has changed my life. I’m sure many of you can say the same. Particularly as I get older and my life gets busier (house, pets, kids, work, etc.), I gain a better appreciation for how ecommerce has allowed me to outsource one aspect of my life: shopping. I don’t hate shopping per se, I just don’t have time for it. And apparently I’m not alone. Now in its fifth year, an annual survey by analytics firm comScore and UPS found that shoppers now make 51% of their purchases online. 2016 marked the first time that consumers have been reported to buy more things online than in stores. And this trend is expected to increase at a rapid pace as more and more traditionally brick and mortar retail businesses come online.
To this end, there are two breeds of ecommerce experiences taking shape. Some are designed to give us quick and easy access to products we know we need quickly. We call this on-demand ecommerce. The other breed tends to be more online retail specific and caters to shoppers who aren’t sure of exactly what they want. We call this discovery. In this post, I’m going to focus on the discovery use case and explore how natural language search represents a real turning point for consumers.
Learning from human interactions
Think back to the last time you bought a couch at a brick and mortar furniture store. You probably had some criteria in mind but it’s unlikely you entered the store with a particular brand or model in mind. You likely had one of the following two experiences. One, you browsed the store and spent some time looking at all of the couches. Or, two, you were met at the door by a salesperson who asked you what you were looking for and guided you to a few couches that met your criteria. Your engagement with the salesperson was likely chock full of natural language descriptions, like “deep seating” to accommodate your lazy weekends or “that are durable enough for pets and kids” because your last couch was destroyed by said beings.
Let’s now imagine how this experience would translate to online. Because of the constraints of traditional interfaces, designers and developers have had to resort to categorizing products using relatively narrow terminology. While it creates a simpler UI, the problem with this approach is that it often fails to organize information in a way that makes sense to all users. And it often fails to present all of the most relevant results to users. Natural language search has the ability to overcome these limitations.
What is natural language search
While it may seem obvious, what we mean by natural language search is a way for users to search for information using their own words. And within the context of a specific application, that experience needs to have an awareness of the types of products available and the terminology used to describe them. One of the biggest challenges is connecting the dots between the user’s choice of words and the products that best fit their description. In this context, when we look to understand the user’s intent, it goes beyond understanding that they want to find a couch to understanding the criteria they use to describe their ideal couch. Herein lies the complexity…
Let’s refer back to our earlier example, and assume the user says something along the lines of the following statement, “I’m looking for a smaller couch that has deep seating and can stand up to the wear and tear of kids and a dog.” The key terms here are “smaller couch”, “deep seating” and “stand up to wear and tear of kids and a dog”. I can tell you right now that no online furniture retailer has a category called “stand up to wear and tear of kids and a dog”. They will likely have some metadata on products that includes “durable” or “stain proof”. And there may be some terminology along these lines included in the filters, product descriptions or reviews.
To address this problem, an effective search solution needs to be able to ingest all available product data (e.g. names, descriptions, tags, reviews, ratings, categories). It also needs to be able to abstract the key terms in that data set to match the user’s description with the right products. Think of it as a virtual salesperson that is all knowing about the products available and is able to match a user’s request to the best products that match their description.
If you’re interested in learning more about Voysis Search, request a demo today!
Get the latest content from Voysis.