How to turn a search experience from bad to good

Search results are often overwhelming. Tone Nordbø suggests a way to make it far easier to find the gems you are looking for.

The quality of online searches has increased significantly in recent years, but websites with a vast amount of content still leave it up to the user to have to go through a whole bunch of irrelevant search results to find the relevant ones. Online stores such as Etsy and eBay have categories and filters for helping users find what they are looking for, but chances are you’ll be spending a lot of time searching, unless you are a ninja at search techniques.

Through spending hours and hours on sites like the ones mentioned above, I realized one night just how far these kinds of searches are from giving a good user experience. I have a suggestion for how to move a big part of the search workload from the user to the system, but first a (girly) real life example to illustrate how this kind of search experience is today from a user’s perspective. If you’ve spent your fair share of Friday nights searching for something on Etsy or eBay and know what it’s like, feel free to jump to the “Suggested solution” part further down the page, I won’t judge.

The problem

I saw a pair of earrings somewhere, and went to Etsy, my go-to earring provider (and favorite website of all!). I didn’t know what to call the shape of the earrings, so I started with what I knew, that they were made of rose gold, and that they were studs. I wrote “rose gold studs” in the search field, and got 4,679 results. In addition to the search phrase, I used the categories to filter my search. I chose “handmade” as item type, and filtered my results through choosing “Jewelry” –> “Earrings” –> “Post” in the categories. I could try to filter by color as well, but in my experience that usually excludes relevant items, and includes a bunch of irrelevant ones.

Filters are great for improving relevance of search results (precision), but often it’s hard to choose which filter to use. For example, the earrings I was looking for could might as well be in the “Metal” or  “Metal work” categories. With using filters, I got 1,182 items, which is a lot of earrings, so I tried to further reduce the number of items to look through (recall). I started scrolling, and noticed that there were several items in yellow gold listed, so I scrolled back to the top again, and added “-yellow” to my search. This reduced the number of results to 1,058 items.


Etsy search results


I continued to try to exclude words from the search result. Every time I scrolled down the page and went on to subsequent pages to see which words occurred often that I could remove, and the number of items decreased a little every time I updated the search. I found it more than a little frustrating that every time I did a new search, I would have to start from the beginning, looking through the items I had now seen many times, and which were not what I was looking for. And for every word I removed, I would have to go through more and more pages to find a new word that occurred often. And of course I did not know which words occurred often in the pages I didn’t go through, so I based the exclusion on what I saw on the pages I did go through. For example, the hassle of excluding the word “horseshoe”, which occurred what seemed like several times in the first few pages, only reduced the result by two items, and left me having to start at the beginning, going through several pages I had already been through all over again.

After excluding a total of seven frequently used words from my search, I gave up trying to reduce the number of items in the search result, as I had to go through too many pages to find a new word to exclude. The search now resulted in 822 items, divided into 20 pages, and I spent the rest of the night looking through all of these, as I wanted to know all my options before choosing which earrings to buy.

With such a vast user group as sites such as Etsy and eBay has, I doubt that all of their users even know about the possibility to remove unwanted words from their search results, and probably spend a lot of time looking through irrelevant items in search of whatever it is they are looking for. Online shopping is supposed to be efficient, but it seems as though every night I look for something on Etsy or a similar site, hours go by. That of course has to do with the fact that Etsy is awesome and has everything, but still, it affects my user experience.

The suggested solution

The goal is to improve the relevance (precision) of the search results through helping the user to reduce recall. Instead of leaving it up to users to look for words that occur frequently in the search results, my suggestion is to provide the users with these words automagically once the initial search has been conducted. With this search pattern, the user can then choose to exclude the words that are not relevant, and leave the ones that are. Stop words would of course have to be excluded automatically, and the words would be listed by number of occurrences (you could even show the number in relation to the word). The user would not have to go through page by page looking for words to exclude, but could remove them easily one by one. This search pattern could be used in a number of sites where the amount of content makes it tricky for the users to find what they are looking for, and significantly improve the search experience. Using Etsy as an example, it could look a little something like this: