Search, via Google and other search engines, is an integral part of our daily online experience, yet when it comes to ecommerce, many store owners hugely underestimate the value of on-site search. I’ve worked with lots of merchants who have seen huge improvements as a result of investing in improving their search – particularly those who work with more complex products and larger catalogues.
All too often, ecommerce stores place much far more emphasis on category-style navigation within their stores, and treat site search as something of an afterthought. Retail giants like Amazon and eBay promote their search function hugely – as it allows users to provide more detail, resulting in a faster and more accurate route to what they’re looking for. At the enterprise-level, retailers are starting to focus increased attention on search, and natural language processing is currently at the forefront of changes in site search technology. John Lewis , Currys and AO.com are great examples of this.
Before looking at what natural language processing means and what benefits it can bring to both store owners and customers, it is worth spending a brief moment examining how standard ecommerce site search can often leave customers feeling disappointed or annoyed.
One of the main problems with many on-site search tools is that the search algorithm works on a simplistic keyword-match basis. For example, a search for ‘mens blue cotton shorts’ might typically return matches for items that contained the words ‘mens’ or ‘blue’ or ‘cotton’ or ‘shorts’, without being able to see that there is a relationship between the words that make up the search term. In the example given, it’s likely that some results won’t even be shorts at all. This is perhaps an extreme example, but it does highlight how easy it is for search to deliver results that are not relevant.
Using natural language processing in ecommerce site search
The demand for natural language processing (or NLP) in ecommerce search is growing rapidly, with larger retailers looking to improve the way they interprete queries and also handle errors and synonyms more effectively. NLP essentially tries to understand the query being searched, rather than just looking at the use of keywords.
Looking in depth at how NLP can achieve these improved search results, let’s take the example of ‘mens blue cotton shorts’ again. An NLP search could interpret the word ‘mens’ to ensure that the store’s faceted, or layered, filter ‘Mens’ was selected, thus returning only men’s clothing. Next, if products have been set up with colour attributes, the colour ‘blue’ would be automatically selected too. So, even if products in the catalog did not actually contain the words ‘mens’ or ‘blue’ in their titles or descriptions, they could still be returned as relevant, if an NLP search solution was in place.
There are many other ways in which NLP search can provide more accurate results within ecommerce stores. By understanding weights and measurements, for example, accuracy can be improved. Since customers use both metric and imperial systems when searching, it’s likely that we will see more and more systems that can take a search input containing size or weight information and match it intelligently by converting, where appropriate, to the measurements used throughout the store. For example, on an ecommerce store selling DIY products, a customer that searches for ‘4 inch screws’ will be presented with screws measuring around 100mm, and not screws that are too small or too large.
The benefits of using NLP-based search
The most obvious benefit to come from an NLP approach to ecommerce search is increased customer satisfaction. Nothing is more infuriating for a customer than to search for something they clearly know is available in the store, and for the dreaded ‘No Results Found’ to be displayed. A patient customer might try again, taking the time to think of an alternative phrase to try. The average customer, however, will simply go back to Google and type the search phrase there, probably finding better results from competitor sites. With improved search results, bounce rates and cart abandonment can both fall significantly, along with increases in conversion rates and revenue.
NLP-based on-site search systems, like Klevu, have been proven to boost sales conversions dramatically, and also to increase average order totals. The ‘search to cart’ ratio is a key metric here, and is defined as the number of items added to the cart, divided by the number of searches performed by the shopper.
The below example, from Zimmermann, shows how different ankle shoes are grouped and served, despite not using the keyword ‘shoes’.
This is just one example of how the product is matched with noun association. NLP can have a really big impact in instances like this, which would’ve previously shown irrelevant results.
Whilst ‘no results found’ is annoying for online shoppers, the reverse situation, where the customer is presented with pages of irrelevant results, is just as bad. By adopting an intelligent NLP search, rather than a keyword-driven approach, irrelevant results are minimised, and the customer finds what they are looking for more effectively.
NLP-based search has become the norm for most search engines, and in fact, we take it for granted when using Google and Facebook. The same semantically-aware technology looks set to become the norm for on-site search, within ecommerce and across other online sectors. Klevu have delivered this technology to a number of enterprise-level merchants, including HongKong, Jack Daniels and many more – you can see how the experience works on the Zimmermann site, by typing things like ‘ankle shoe’.
Other opportunities for search improvement
A lot of merchants who are investing in search are also using self-learning technology, which helps to optimise the products that are being served and how they’re being merchandised automatically. When used alongside an NLP solution, you can expect significantly more effective results from search due to further increased accuracy and more relevant results.
Another potential failure in standard site search is the lack of integration between the product catalog and other areas of the site, such as FAQ pages or information pages, such as delivery. Sometimes a customer may wish to search for information, rather than for products, and a product-only approach to search usually means that this is not possible.
Paul Rogers is a consultant for search company, Klevu.
You are in: Home » Guest Comment » GUEST COMMENT The benefits of natural language processing for ecommerce site search
GUEST COMMENT The benefits of natural language processing for ecommerce site search
by Paul Rogers
This is an archived article - we have removed images and other assets but have left the text unchanged for your reference
Search, via Google and other search engines, is an integral part of our daily online experience, yet when it comes to ecommerce, many store owners hugely underestimate the value of on-site search. I’ve worked with lots of merchants who have seen huge improvements as a result of investing in improving their search – particularly those who work with more complex products and larger catalogues.
All too often, ecommerce stores place much far more emphasis on category-style navigation within their stores, and treat site search as something of an afterthought. Retail giants like Amazon and eBay promote their search function hugely – as it allows users to provide more detail, resulting in a faster and more accurate route to what they’re looking for. At the enterprise-level, retailers are starting to focus increased attention on search, and natural language processing is currently at the forefront of changes in site search technology. John Lewis , Currys and AO.com are great examples of this.
Before looking at what natural language processing means and what benefits it can bring to both store owners and customers, it is worth spending a brief moment examining how standard ecommerce site search can often leave customers feeling disappointed or annoyed.
One of the main problems with many on-site search tools is that the search algorithm works on a simplistic keyword-match basis. For example, a search for ‘mens blue cotton shorts’ might typically return matches for items that contained the words ‘mens’ or ‘blue’ or ‘cotton’ or ‘shorts’, without being able to see that there is a relationship between the words that make up the search term. In the example given, it’s likely that some results won’t even be shorts at all. This is perhaps an extreme example, but it does highlight how easy it is for search to deliver results that are not relevant.
Using natural language processing in ecommerce site search
The demand for natural language processing (or NLP) in ecommerce search is growing rapidly, with larger retailers looking to improve the way they interprete queries and also handle errors and synonyms more effectively. NLP essentially tries to understand the query being searched, rather than just looking at the use of keywords.
Looking in depth at how NLP can achieve these improved search results, let’s take the example of ‘mens blue cotton shorts’ again. An NLP search could interpret the word ‘mens’ to ensure that the store’s faceted, or layered, filter ‘Mens’ was selected, thus returning only men’s clothing. Next, if products have been set up with colour attributes, the colour ‘blue’ would be automatically selected too. So, even if products in the catalog did not actually contain the words ‘mens’ or ‘blue’ in their titles or descriptions, they could still be returned as relevant, if an NLP search solution was in place.
There are many other ways in which NLP search can provide more accurate results within ecommerce stores. By understanding weights and measurements, for example, accuracy can be improved. Since customers use both metric and imperial systems when searching, it’s likely that we will see more and more systems that can take a search input containing size or weight information and match it intelligently by converting, where appropriate, to the measurements used throughout the store. For example, on an ecommerce store selling DIY products, a customer that searches for ‘4 inch screws’ will be presented with screws measuring around 100mm, and not screws that are too small or too large.
The benefits of using NLP-based search
The most obvious benefit to come from an NLP approach to ecommerce search is increased customer satisfaction. Nothing is more infuriating for a customer than to search for something they clearly know is available in the store, and for the dreaded ‘No Results Found’ to be displayed. A patient customer might try again, taking the time to think of an alternative phrase to try. The average customer, however, will simply go back to Google and type the search phrase there, probably finding better results from competitor sites. With improved search results, bounce rates and cart abandonment can both fall significantly, along with increases in conversion rates and revenue.
NLP-based on-site search systems, like Klevu, have been proven to boost sales conversions dramatically, and also to increase average order totals. The ‘search to cart’ ratio is a key metric here, and is defined as the number of items added to the cart, divided by the number of searches performed by the shopper.
The below example, from Zimmermann, shows how different ankle shoes are grouped and served, despite not using the keyword ‘shoes’.
This is just one example of how the product is matched with noun association. NLP can have a really big impact in instances like this, which would’ve previously shown irrelevant results.
Whilst ‘no results found’ is annoying for online shoppers, the reverse situation, where the customer is presented with pages of irrelevant results, is just as bad. By adopting an intelligent NLP search, rather than a keyword-driven approach, irrelevant results are minimised, and the customer finds what they are looking for more effectively.
NLP-based search has become the norm for most search engines, and in fact, we take it for granted when using Google and Facebook. The same semantically-aware technology looks set to become the norm for on-site search, within ecommerce and across other online sectors. Klevu have delivered this technology to a number of enterprise-level merchants, including HongKong, Jack Daniels and many more – you can see how the experience works on the Zimmermann site, by typing things like ‘ankle shoe’.
Other opportunities for search improvement
A lot of merchants who are investing in search are also using self-learning technology, which helps to optimise the products that are being served and how they’re being merchandised automatically. When used alongside an NLP solution, you can expect significantly more effective results from search due to further increased accuracy and more relevant results.
Another potential failure in standard site search is the lack of integration between the product catalog and other areas of the site, such as FAQ pages or information pages, such as delivery. Sometimes a customer may wish to search for information, rather than for products, and a product-only approach to search usually means that this is not possible.
Paul Rogers is a consultant for search company, Klevu.
Read More
You may also like
Register for Newsletter
Receive 3 newsletters per week
Gain access to all Top500 research
Personalise your experience on IR.net