ecommerce search relevancy

Posted by Peter Curran

Today we’re introducing a series of blog posts about ecommerce search recall and relevancy.  Other companies, notably the Baymard Institute, have written informatively about the different kinds of queries that searchers perform and the critical role search plays on any ecommerce site. Their work helps us think about users’ goals systematically and map available technology to these goals when designing ecommerce search solutions in order to drive better search results and consequently positively impact desired outcomes including conversion, average order size, customer engagement and brand experience.

Ecommerce search, done right, is a critical foundation to overcoming challenges that ultimately impact revenue. Considering the time and expense to drive consumers to your site, even if every part of your business is stellar – including the products or service – poor search can lose customers well before they get to the cart.

Our blog series will focus on how many ecommerce search solution designs fall short. If Baymard Institute’s work is Psychology 101, ours is the Abnormal Psychology class that follows.  We’ll group ecommerce search abnormalities into “problem themes” and focus on a new theme with each post.

Fixing One Problem Can Fix Many Problems

Grouping problems into themes is useful because, typically, a single failing query fails in a way that is repeatable across many other searches.  For example, a search for a sofa on a department store site may fail in a way that is the same as a search for coffee.  Changing how the search engine finds and orders matches for sofa may also fix the search results for coffee and a myriad of other related searches that might otherwise go unnoticed.

Problem Theme 1: Search results are tertiary, not primary

Your search delivers something related, but misunderstands the customer’s intention.

Accessorize – but only when it makes sense. Several years ago, I bought a Nespresso coffee maker so that I could have cappuccinos at home.  I got the maker and the frother and while I knew I’d have to buy pods, I didn’t realize the depth of the hate I’d feel for Nestle every time I had to order more.  And yet, I have to give it to them: accessories make money. The machine is only a drop in the cup, as it were, compared to the pods business.  Unfortunately, in the onsite search experience for many companies, it can sometimes seem that accessories are all that a company sells.

For example, suppose a search engine is configured to match a term to the product titles (very common) to determine if the product should be recalled.  Now suppose you have two products, a “Vacuum Cleaner” and a “Vacuum Bag” and the user searches for “vacuum”.  How can the search engine know that the Vacuum Cleaner product is likely more relevant than the Vacuum Bag product in a search for vacuum?  The answer is, with no other information, it can’t.

Many major sites we’ve tested handled this problem well when the ecommerce search was for a single term.  Searches for bed did not generally return sheets, searches for vacuum did not generally return vacuum bags, and searches for crib did not generally return bumpers. Of course, the problem can go the other direction too – it makes sense for a search for “coffee” to recall coffee makers, but those products should probably be beneath actual coffee.  This suggests that the sites have either manually tuned the results for these simple search terms or indexed other data that allows them to discriminate between products and accessories and help the technology understand the concepts.

Even with these sites where search tuning is slightly more advanced, it’s possible to issue a query that yields this very common problem.  Take the phrase “king size bed”.  Many companies – including Restoration Hardware, Pottery Barn, Sears, Amazon and Crate and Barrel – handle this query well, but Target reveals the issue when mostlyaccessories are returned instead of the furniture items the shopper obviously desires.

Unfortunately for the Target shopper, this search experience for “king size bed” is unfruitful (and likely irritating). Including furniture and king-size as available guided navigation refinements provides some mitigation, but the customer has very clearly stated what they are looking for and the site is still making them work to locate the item. Even with these refinements, they are not exposed to the entire selection of beds available. In comparison, the products available by browsing the site categories to the king bed section shows more than 350 relevant products – none of them accessories. The site has forced the customer to work hard to see a more limited selection which may not even land them in the right place. They may be leaving in frustration rather than already in the checkout.

The intention is not to call out Target specifically (many sites have this issue), nor is this an indictment of their technology platform. We’ve often seen clients that have a successful strategy for curating product results when customers browse in categories, but are unable to take advantage of that in (or apply similar effort towards) search. This illustrates a search problem that should be improved from a best practices viewpoint and can be improved with correct configuration and tuning.

Analyzing Your Own Search

So, how do you know if your ecommerce site has similar issues? The simplest way is to compare the product assortment that results from a search to the assortment shown when you browse to the equivalent category. If the search and browse results are very different consider whether the assortment in the category view is also what you would prefer to see for search, and why. Keep in mind that your customers may be expecting different things from the two experiences – they may choose the browse path because they are hoping to discover and the search path because they are looking for something specific (or the other way around – it varies depending on what you are selling). It’s likely that the relevant results should be different based on their choice of interactions. Should you show them a broader assortment of products and accessories when they search, when they browse, or should it be the same?


Return the result the customer wants – quickly and efficiently

We’ve covered one type of issue that is commonly found in un-tuned search implementations, regardless of technology platform. At the most basic level, it is often one of the most frustrating experiences from the customer’s viewpoint. Luckily, if you find problems that follow these patterns on your ecommerce website, they can be fixed by using (or enhancing) the ecommerce search technology you currently utilize.

Dive deeper into how relevance tuning is done.

Need further strategic or technical help?