Search Query Analysis – Part II

Search Query Analysis – Part II

Today, shoppers are becoming more accustomed to the robust features of major web and social media search engines. These search engines have the uncanny ability to interpret and yield relevant results to complex search queries intelligently; these expectations of the shoppers are carried over to e-commerce searches as well.

In the last report, we explored the 4 different kinds of search queries and performances: exact searches, product type queries, non-product queries and symptom search on the Australian eCommerce market. In this part of our report, we’ll explore the concepts of “feature and thematic searches”. We’ll deep dive into their use cases and talk about how strong Product DNA can be a turning point for efficient onsite search functionality. 

Click here to read the first part of the report.

Feature Search

Often shopper will a have predefined notion of what they want the product to have, more often than not, these criteria relate to features of a product.  A  feature search occurs when a shopper includes one or more features/elements in their search query that they would like the product to have. For example, a shopper might not want just any “dress” but will often be looking for something slightly more specific such as a “red dress”.

The term “Features” should be interpreted in a broad sense, referring to any product attribute. Features could thus be a product’s colour (“red dresses”), material (“silk”), drive”), or length (“knee-length”), price (“$100-$200 “), brand (“Kate Spade”), or size (“size 8”), the list goes on. It is important that all significant product attributes be searchable.

Generally “Feature” searches are almost always used as a qualifier for different search type — a way to filter the results of a specific search. For instance, a user may perform a “Product Type” or an “Exact search” and combine it with a “Feature” search, only to get a subset of those products (e.g., “blue maxi dress” or “faux fur jacket”). Shoppers also consistently combine “Feature” search with “Thematic”, “Symptom”, and “Compatibility” searches.

Irrelevant Search results for the search term “Faux Fur Jacket”

Feature Searches by far are the most common search query, which makes it a definite “must-have”. Sadly, while 94% most eCommerce websites responded to an elementary feature (red dress, blue coat) search when the features became slightly tricky, the performance dropped quickly. Amongst the top 100, only 19% responded to the search query “silk skirts”, and only 8% of the website responded to the query when the product attribute “shimmery or sequin” was mentioned. 

Accurate product tagging can yield useful results.

Today, shoppers are becoming more accustomed to the robust features of major web and social media search engines. These search engines have the uncanny ability to interpret and yield relevant results to complex search queries intelligently; these expectations of the shoppers are carried over to e-commerce searches as well.

The ideal solution for this would be to dynamically apply any features searched for as filters on the results page. This will increases transparency and boost the shoppers’ control – with the user being able to see what is and isn’t included, and be able to toggle related filters on/off quickly. For instance, in the example mentioned above, the “red” aspect of the “red dress” query is applied as a colour filter, with the user being able to see and toggle the other available colours of the dress. However, more complex feature search queries can be enhanced only if the eCommerce retailers knows his product DNA well, what does that mean – awareness of the products attributes, types, style and usage very well. All of these are potential make up a part or type of search query, therefore tagging your products correctly and creating relevant filters based on those tags can help retailers bring their shopper to the product they need the most. Learn more about tagging here. Read more about product tagging here: Benefits Of Automated Product Tagging

Thematic Search

What exactly constitutes a “boho style”, or “athleisure”, or a “retro dress”? These concepts are easily recognizable by very difficult to define. Especially in an eCommerce search context, setting their exact boundaries can be very challenging.  

Categorizing “boho style”

#6 Thematic Searches are often complex to define because they are inherently vague. Often these searches include indefinite boundaries of usage locations (e.g., “living room furniture”), environmental or seasonal conditions (e.g., “spring jacket”, “boho style“), events and occasions (e.g., “evening gown”), or even promotional attributes (“sale top”). 

Irrespective of their ambiguity, they are genuine concepts to shoppers who consistently use them in their search queries when shopping specifically for apparel and furniture.

Being able to implement or support “thematic” search a great deal of interpretation is required, both in terms of the meaning of the actual query itself and most importantly in the internal tagging of products. It is vital that a query for e.g. “retro-style” presents all the relevant products and not just those products which happen to have these keywords in their title or description. Accomplishing this requires thematic tagging of the product catalogue to determine, e.g. which outfit or item would be suitable for retro style and which wouldn’t. Credits: Amazon.Com

The ideal support for most “Thematic Searches” is often achieved by having these themes as actual filtering values. These values can then be pre-applied when users search using these terms; it will give your shopper direct insights into how your eCommerce site has interpreted their thematic query and an easy way to narrow or broaden their thematic search. An added advantage is that this will also improve the general navigation based filtering experience significantly. However, thematic tagging of a catalogue can be a herculean task, which is why automating these processes can prove to be highly advantages for any eCommerce retailer. Check out the benefits of auto-tagging here. 

Let’s take a look at another example: a search for a “cocktail gown” requires products to be grouped by an occasion (a social concept). Similar mappings may be necessary for abstract queries such as “boho style” — an actual search query we tested. Only, 56% of 100 eCommerce website responded with practically broken search listing, i.e., they were marginally better than other non-responsive websites. Out of 56% that responded to a thematic search showed a sparse list of the product. However, upon close inspection, there were several items in the catalogue that could have been listed as “boho style”. Since shoppers sometimes think in these “Thematic” terms when typing into the search field, their product exploration is disrupted when results don’t meet their expectations — forcing shoppers to either refactor their queries or abandon website.  

If an eCommerce website doesn’t support “Thematic” search queries, shoppers are left with few, absent, or irrelevant results. This costs shoppers more time to rethink (and retype) query wording and often leaves the impression that the products sought simply aren’t available. In fact, 44% of our benchmarked sites have problems handling “Thematic” search queries, if the thematic identifier doesn’t happen to be part of the product title.


Feature or thematic type search queries must be supported in an eCommerce environment, primarily because the fashion and apparel industry is very trend-driven. Products largely remain the same ut may often be re-hashed and rebranded and reintroduced in the market. Therefore, it becomes vital to keep up with the trend and optimize search for maximum discoverability.  

Know Your Product DNA? 

You product DNA is length and breadth of your products makings, i.e., its attributes. In order to efficiently map your product to the relevant categories and optimize it for your search, an eCommerce retailer must know the attributes of each product and know it well. 

As mentioned above and in the past reports, shopper search in a different context when it comes to eCommerce to provide them access to your products tagging your products correctly becomes crucial. As tedious or repetitive the activity it maybe – it does yield significant results when implemented correctly. It helps boost search capabilities of the website, helps create filters, and categorize products faster and better. Today tasks like these can be automated and not longer are time-consuming and tedious. Save on time, cost and effort with the help of Okkular Tag-Gen. Our AI is trained to extract out product attributes in a matter of seconds; now you can tag products at lightning speed, launch products to your catalogue with a click of a button and create unlimited product filters. 

About Us:

Okkular is a Melbourne based Australian tech startup. Okkular’s vision is harnessing Deep learning driven artificial intelligence to help retailers improve customer experience, reduce costs and drive sales with insightful automation and personalisation. Our solutions have been developed by having conversations with multiple fashion retailers and marketplaces.

Team Okkular has given significant importance to understand fashion business pain points, use-cases and learnt from their feedback.
Co-creation is at the heart of Okkular’s vision and values. Okkular has been seed funded by angel investment and also received a recent innovation grant of $200k from the Australian Government’s Department of Industry, Innovation and Science [Part of the Entrepreneurs’ Programme]. This grant will help Okkular run further Pilots/POCs and commercialize the offering promptly.

Wanna know more about our offerings head to

Wanna schedule a demo: