The Art of E-Commerce Site Search

Natural Language Processing allows users to search the way they speak by computing the overall meaning of a search query, instead of individual keywords. NLP helps ensure that users find what they're looking for, regardless of using incomplete, ambiguous, or unstructured questions. It's like having a human agent receive a query and return relevant results.

Providing an effortless shopping experience for customers is a key driver to the success of any e-commerce business. E-commerce search is a big factor in this experience and when done right can provide a significant competitive advantage.

An effective on-site search experience leads to higher sales and increased customer loyalty. A substandard search experience will increase bounce rates, cart abandonment, and can lead to the permanent loss of a customer. A recent large-scale usability study benchmarking the search experience of the 50 top-grossing U.S. e-commerce web sites revealed a significant missed opportunity for most e-tailers.

When it comes to e-commerce search, most web sites focus on SKU’s (stock keeping units) to manage their catalogs because product data is a logical approach to navigation. The problem with that is users don’t normally query for SKU’s. Instead, they search for products and services by describing attributes, features and characteristics; such as price, size, color, voltage, dates, location, etc. This can lead to incoherent search results, or worse — zero results — when in fact the item is actually available.

A particularly frustrating example I recently experienced happened during a search at a well-known online toy retailer. I was shopping for something I knew they sold, a Lego Death Star. Needless to say I was surprised when my search for “lego death star under $100” came up empty. Not only that, the search mechanism tried to correct me with a “Did You Mean depth” error message.

Ineffective e-commerce search should never be the reason for a lost sale. Instead, it should be an opportunity for conversion. Let’s explore several ways that e-commerce search can work harder for the bottom line.

Natural Language Processing

Today there are more online shoppers than ever, and equally as much competition among retailers. Despite the huge amounts of money spent on marketing, sites can easily lose customers with a single misstep.

Natural Language Processing (NLP) allows users to search the way they speak by computing the overall meaning of the search query instead of individual keywords. It helps ensure that users find what they’re searching for, regardless of using incomplete, ambiguous, unstructured questions. It’s comparable to having a human agent receive a query and return relevant results.

Here is an example of search results not matching the request at Amazon:

Let’s say a musician is looking for a new bass guitar. The standard design for an electric bass guitar has four strings, although five strings is also a common option. Notice that the search queries are basically the same:5 string electric bass guitar above $1,000 Bass guitar with five strings for $1,000…but results are very different:

The problem is that most search engines are keyword based. Using this retrieval approach, many text indexing systems will only pick up and process every word in a string, except for commonly occurring stop words that are considered unnecessary noise.

Keyword searches also have a difficult time distinguishing between homographs, or words that are spelled the same way but have different meanings (i.e. lead, meaning to direct someone or something vs. lead, the element). This often ends in frustration for the user, because this type of search engine is unable to discern any meaning and often delivers zero results, or produces results that are completely irrelevant to a user’s query.

A search engine with NLP capability will yield more accurate results than a basic keyword engine.

Long Tail Search

Search queries with three or more words are considered “Long Tail.” Whileonline marketers invest a considerable amount of time and money trying to find the keywords that will deliver ROI, they also pursue highly specific three-or-four-word phrases, because those are more likely to convert a targeted buyer. Typically, the longer and more detailed a search term is, the more likely the customer is to purchase. However, most e-commerce searches do not support more than a few words — essentially alienating serious buyers.

This is where intelligent search comes into play. Natural language search understands long tail and complex queries that might include idioms, slang, synonyms, misspellings and abbreviations. Equipped with NLP technology, e-commerce search can manage unstructured queries, such as:

  • screwdrivers 6 inches
  • car radio with bluetooth compatible with iphone
  • taylor swift concert in san francisco bay area

In these cases, [popular product name] would be any combination of words that describe the product, and features the set of attributes required to fulfill…price, size, color, etc.

Well-Organized Categories and Faceted Search

Faceted search, or guided navigation, is the standard for categorizing modern e-commerce sites. Even so, only 40 percent of web sites today use faceted search, despite it being the foundation of contextual filters. This model allows customers to refine or navigate through products by checking off attributes, or facets, such as ‘Women’s Clothing’ and/or ‘Dresses’, which are then further narrowed by choosing size, color, or price options.

Notably, what many sites fail to do is link their search systems with their facets. Through linking, facets are automatically filled out. For example, a search for black bag under $50 for a woman would select the facets ‘black’, ‘under $50′ and female’ in the left-hand column. Giving shoppers the ability to filter and sort through products helps them find what they want, even if they are not completely sure what they are looking for.

Make Sure Search Understands Symbols

Referring back to the search for the Death Star toy under $100, notice that the U.S. dollar symbol ($) was removed during the search. But why? That symbol was an important part of the query.

Search functions should be able to understand metrics in all of the different ways a customer could type them. For example: IN, inch, in., and ” all mean the same thing.

Natural language search does not get tripped up by units of measurement, characters, symbols, web site jargon or product models. NLP can also interpret calculations; for example, someone searching for a 5 gallon paint bucket would also be presented a search result of 20 quarts paint bucket, because they are the same amount or size.

Eye-Catching Search Bar

The search bar is similar to the warm welcome someone gets from a storeowner when they walk in the door. It represents an offer of help if the customer wants, or needs, some guidance through the store.

The largest online retailers understand the importance of a conspicuous search bar in their design, yet it is a feature that many online stores overlook.

Most visitors to stores are driven there for a particular reason, and the search bar gets them to their goal faster than browsing.

Companies that do a good job of this often see customers using the search box as their very first action — especially if they are looking for something specific. Having a prominent search bar can mean the difference in completing a conversion.

Jordi Torras is CEO of Inbenta.


  • Other important factors for search also include predictive or suggested search. You will be surprised how many people can’t spell from looking at search analytics. The other is product discovery during search, so recommending products as shoppers type can greatly increase engagement and conversions.

Leave a Comment

Please sign in to post or reply to a comment. New users create a free account.

Related Stories

E-Commerce Times Channels