Semantic search can grasp the intent of the consumer, but it requires a retailer to work on the content of its product catalog.

On-site search technology has seen a lot of innovation over the last few years, with more and more providers offering features around self-learning capabilities, behaviour-based merchandising and of course faster results. The majority of solutions, still, however, are keyword-matching search engines with the possibility of handling typos and spelling errors.  At the enterprise-level of the market (i.e. large and ultra-large online shops) there’s never been more demand for UX-orientated features capable of getting users to what they’re looking for more efficiently. However, even understanding consumer intent in its natural form is still beyond the reach of most search engines.

There’s plenty of research highlighting that users who complete a search are considerably more likely to convert and this is something we’ve seen with lots of our customers. In fact, two of the most common reasons why merchants start conversations with us is as a result of analysis that has highlighted the need for better results or they’re investing as a result of seeing the higher value of users who are searching. 

The statistics surrounding site search in online retail are quite telling in terms of how few ecommerce stores are truly capitalising on the power of their search functionality. For example, according to this piece from Smash Magazine, only 40% of websites have faceted search, 70% require users to input the exact keyword and 60% of stores are unable to understand queries with symbols or abbreviations.

Outlined below are some of the most important elements to bear in mind when using your site search functionality to drive an increase in sales:

Optimal language processing is key


A lot of the larger, more proactive online retailers are already investing in technology that can handle queries more effectively—particularly when it comes to error handling and sentiment analysis.

Klevu, as well as some of the other solutions available, is able to match queries based on more than just metadata, which allows for much better query matching. (I have served as an advisor for Klevu.) This technology also pulls in additional data to allow users to search for things like “most popular armchairs,” which is an interesting behavioural trend that Klevu has seen with some of its larger customers.

Addressing search queries in their natural form, or, semantic search, as it is termed, requires a lot more than understanding alternative nouns or adjectives. Semantic search starts to show its real power in understanding the intent of the consumer. Understanding the intent requires a lot more enrichment of the catalog and capability to process the search query in real time. These two capabilities will enable the returning of relevant results for a search query such as ‘executive chairs around 500 dollars’. It should be noted, however, that the top 50 grossing U.S. retailers are still largely missing such capabilities, as reported in the analysis of search from Baymard Institute.    

Make your search do the work

A new and emerging method of auto-suggest (auto-suggest being when results are shown before a query is completed, e.g. displaying Nike products when “Ni…” is entered) seems to be showing products and facets—helping to speed up the overall search experience for users. One key recommendation I would make if you’re implementing auto-suggest though, is that you ensure that you’re still able to record the queries being used. For example, if you’re using Google Analytics and one of the third-party solutions available, there’s a good chance the queries won’t be reported in the search reports, as you’re not actually completing the search. You should be able to record these queries by changing the parameter in site search settings in Google Analytics (this will differ depending on the provider).


The above example is from

It is important to design your auto-suggest search experience based on careful analysis of the shopping segment, demographics and consumer segment. For example, in the nursery or fashion categories, it is more important to show intuitive keyword suggestions and categories to let the shoppers discover more on landing pages, whereas in other categories it could impact the user journey. This is also something that should be measured regularly and optimised.

The above example is from

Optimise your results!

Merchandising your search results is key—particularly if a high proportion of your users are completing searches. Most of the search tools available today allow you to create rules to promote key products within set queries, however there is a clear need for a more intuitive solution.


Results should really be auto-merchandised based on how they user is responding to different products – which is something that the top tier of search solutions do. This is actually one of the primary reasons why merchants tend to use third party search technology, instead of the out of the box search within ecommerce platforms.

In addition to auto-merchandising, which promotes the products that get getting the best click-through rates etc. —you should also be promoting the products that drive the best margins or you’re looking to sell etc.

Make sure you know what’s working and what isn’t

Search represents a treasure trove of valuable information as it provides a real insight into how consumers are engaging with your site. We published this piece which highlights some of the things that merchants should be reporting on around search, but here are a few metrics I’d suggest keeping on top of:

  • Most popular search queries—this is important as it highlights potential issues with the website and also provides general insight into how users are using your website. This data can also be useful for buying decisions.
  • Pages where the most searches are initiated—again, can highlight issues with navigation or merchandising.
  • Highest revenue generating search queries—Can highlight a need for new pages or navigational changes. Also useful for merchandising search results.
  • Conversion rate and AOV for journeys that include a search vs journeys that don’—can help you make the decision on how much you should promote your search function.
  • 0 result searches—Important to understand and fix (be it with custom messages, serving different products or redirects)
  • Percentage of users using search—Again, can be indicative of navigation issues and it’s important to understand peaks.
  • Average number of searches per session that include a search—Helps to understand the overall performance of search.

There are just a few metrics and data points that can be monitored, but most of them should really be measured in some way. We generally recommend that our customers measure these kinds of things on a monthly basis.


To conclude, is a vital part of your customer experience and it’s really important that you invest in providing the best possible results to your users. Improving the results being served and they way that you’re serving them is a proven way to generate incremental revenue through improvements in conversion rates and higher AOV and it’s something that can be done without huge overheads.

Paul Rogers provides enterprise-level Magento consulting and auditing services.