6 Practical Tips For Improving eCommerce Site Search eCommerce business owners are familiar with the power or on-site search, but fail to maximize its potential in their own sites.

For many of the global ecommerce behemoths, such as Amazon and eBay, on-site search has become the de facto method for customers to find what they want to buy. Whilst these two online giants do still offer a category-based navigation system, by far the most popular method for finding what we want on these sites is via search. It can come as a surprise then, that many ecommerce business owners are familiar with the power or on-site search, but fail to maximise its potential in their own sites.

On-site search is such a powerful tool for eCommerce optimisation, just waiting to be used to its full potential. This article looks at assessing site search performance and identifying ways in which that performance can be maximised.

Make sure that search performance is measured

Just like any other development or enhancement on an eCommerce site, analytics should be at the core of every decision. Unless you know how your store is performing, and specifically, how your site search is performing, any changes you make will not be based on factual evidence. Google Analytics provides the most obvious starting point, in terms of data sources. The GA Site Search reports contain a wealth of useful data, to help you understand how your customers are currently using search on site. For example, by comparing conversion rates for search-facilitated orders before and after any design changes, you can ensure that all changes have a positive effect.

Utilise natural language processing

Natural language processing (NLP) is making a big impression in the eCommerce sector currently. NLP offers a much more intelligent and intuitive approach to search. Instead of seeing a search phrase simply as a sequence of keywords, it instead looks to evaluate the actual meaning behind the phrase. NLP strives to actually understand what it is that we are asking, when we search a site. This allows an NLP-driven store to deliver search results that are conceptually related to the search phrase, rather than results that contain all of the individual words making up the phrase. Executed well, an NLP-driven search engine should easily beat a standard search, in terms of relevant results.

A good example of this can be seen on Zimmermann - searching for "ankle shoes", you can see below that the products being returned don't actually mention shoes, but NLP understands that these are types of shoes.

By allowing customers to express their search requirements in a natural way, NLP offers improved customer satisfaction, as well as better overall delivery of relevant results.

Merchandise your results

Many eCommerce stores seem to do little in terms of merchandising their search results. Results can often appear in a random, rather eclectic jumble, spanning multiple product categories and with no real sense of cohesion or relevance.

By checking search term reports to identify key searches, store owners can then perform the search themselves to see if the actual results match what they expect to be shown for those terms. It may be appropriate to redirect certain very specific search phrases straight to a product category, rather than showing the actual search results. A good example of this might be a fashion retailer, who chooses to redirect any searches for "mens coats' to the Men's Coats category of their store, rather than displaying the standard search results for that term.

Another possibility for improving the display of search results is to add some form of ranking or ordering to the results. More advanced solutions, like Klevu, may be able to assess add-to-cart statistics and conversion rates to identify which products to place at the top of the search results. At a more basic level, units sold or page view counts could be driving rankings, so that the most popular products appeared first in the results set. Retailers think nothing of presenting a coherently ordered product category, but often overlook the potential of doing the same for search results.

Self-learning search solutions, like Klevu, will automatically merchandise results based on what users have clicked and bought from different queries - this can ad a huge amount of value and improve the conversion rate for key queries. This can be a highly impactful addition to on-site search for merchants and is something that larger retailers should be using.

Promote the search box

Again, online retailers are often quick to allocate large proportions of their CX / trading budget to page template changes or merchandising software but don't see the value of investing in site search. A poorly-designed search bar can lower search usage dramatically and there are lots of reports around that highlight conversion multipliers for users who complete a search.

Simply making the search bar and search button larger and more prominent is a good way to drive more usage and encourage users to search. If we look again at the giants mentioned earlier, Amazon and eBay, we can see that their search bars take centre stage at the top of the screen, as they understand the need for it and the value it can provide to customers. Other major retailers such as B&Q, John Lewis and Screwfix, all feature strong, clear and central search bars that encourage customers to use them. In contrast, someone like 21 Diamonds don't promote search at all, making things hard for users unable to find what they're looking for.

Any site element that causes a customer to think too much, or to become confused or frustrated, needs to be addressed. This is particularly true of search because it is the most direct channel through which customers can interact with your site, to get what they want. A site search that is difficult to find, offers a poor user experience or that generates seemingly random results can do significant damage to both sales conversion rates and overall brand approval.

Rather than seeing the search function as something that has to be squeezed into page design, as unobtrusively as possible, retailers should consider how they can design their page layout to fit around a prominent search function, actively encouraging shoppers to use it as the preferred form of navigation.

Another element of search that is often poorly handled is the "No results' page. In an ideal world of course, this page would never appear. However, if a search query returns no results, it's important not to abandon the customer with a terse "No results' message, or to suggest that perhaps the customer has made a mistake. I'd recommend trying to add product recommendations and category links to 0 result pages, ideally with personalised listings based on what the user has been looking at etc.

Allow users to refine search results

Layered navigation, or faceted search, is another way to add value to on-site search, by allowing customers to filter search results further, in order to narrow down their search. Screwfix for example, allow customers to filter search results by brand, colour, price, star rating and more.

Layered navigation can improve the user experience significantly, especially for product catalogs containing a very large number of SKUs, or catalogs with a large quantity of product variants, such as colour and size.

A sophisticated NLP-based search could take faceted search a step further, by pre-selecting certain filters based on the search words used. For example, if the Screwfix search above was for "mixer taps less than £100', then the price filter could be pre-set when displaying the results, so that only those products that were actually less than £100 would be shown.

Facilitate for informational searches

Consider, now, the other example we used to highlight the shortcomings of standard search - the lack of integration between the product catalog and other areas of the site, such as the blog or FAQ pages. An enterprise-level search engine provides a more integrated, holistic approach and returns the right information to the customer at the right time. As an example, it is reasonable to assume that if a customer searches an eCommerce store for "delivery costs', that customer wants to know how much it will cost for delivery to their address. They are most definitely not looking for a product that has "delivery costs' as part of its product name or description. If we type that search phrase into Hongkong's website search (one of Finland's largest retailers), we can see an intelligent result, with the search listing all of the pages that provide information around delivery.

Conclusion

From simple design changes to promote the use of on-site search to more sophisticated enhancements to fundamentally change how search works within an eCommerce store, there is much that can be done in this area to dramatically improve the user experience and to increase shopper engagement and sales conversions. With careful analysis of traffic and behaviour, it should be clear which changes are having a positive effect and which need further refinement. As with anything in the world of eCommerce, improving site search should be an ongoing and incremental task. It should, however, definitely be a priority, since the rewards to be gained could be impressive.