How to increase your Ecommerce Conversion Rate – Product Search
In this series, we explain how retail stores can boost ecommerce conversion rates. In this part, we show how to build product search to help your customers discover products quickly and easily.
Throughout this series, we are sharing the definitive best practices that will improve your e-commerce website and increase your revenue and conversion rates. In part 1, we looked at how to effectively use landing pages to drive sales, as this is where your customers will begin their online interaction with you.
Part 2 of the series explains how to extend your search function to ensure your customers can get to correct and relevant product information as quickly as possible. The best way to achieve this is to use ‘search’ to its full potential.
Back in 2004, Google introduced predictive search with auto-complete and this has since become a universal feature that customers expect as part of the ecommerce shopping experience.
Fast, predictive search
A customer’s shopping experience is a journey. This means that the search function also needs to be about discovery, displaying something to them that encourages further exploration and gives them instant access. All this can be delivered by using predictive search and the auto-complete function, which are both elements that we used when designing the ecommerce website for global retailer, Graham & Brown.
Display relevant items, collections or articles
We advocate using predictive search on your site, as it allows the customer to see highly relevant suggestions of products, collections or articles as they type. This gets the customer to the right information as quickly as possible. Predictive search is a global best practice, especially for e-commerce. Usually it takes the form of a drop-down box that pops up immediately with auto-complete suggestions as users type, without them needing to hit the ‘search’ button in order to see typical search queries.
This immediate display gives visitors an efficient way of navigating to relevant results, rather than having to type a search expression, which might be a slightly different term or phrase that will yield a limited number of results.
After incorporating predictive search for Graham & Brown, we saw the percentage of visitors who viewed at least one product page increase by over 20%. It also increased the average time spent on the site by 11.27%.
Include collection pages with suggestions
Including collection pages in your predictive search suggestions is an added benefit to the customer experience. This enables the user to engage further with other items and content that are related to their original search. For example, search for ‘Kelly Hoppen’ on the Graham & Brown site, you will be offered a mixture of relevant products, collection pages and articles.
We also recommend your search functions work across any languages your site caters for, as well as for any products, collections or lifestyle content.
After launching our updated predictive search for Graham & Brown, we saw time on site increase by 16.56% for those visitors who used the search function. In addition to this, we found that the search was providing a much better way for customers to find exactly what they were looking for, regardless of where they entered the site, with the bounce rate decreasing by 35.68%.
Introduce auto-complete function
We recommend you incorporate the auto-complete function in your predictive search. This feature also helps save your customer typing time. This is especially important for mobile or tablet users without a physical keyboard.
Auto-complete presents your customer with the most relevant search matches across your site. This is beneficial because they may not be able to remember an exact title or product line of, for example, a product you sell. This problem can be solved with auto-complete, as search suggestions will help guide them to what it is they are searching for, despite not knowing the exact terms used across your site. They can select the item through visual recall, identifying the product they were looking for out of the auto-complete suggestions.
Help customers save time
Another important advantage of using predictive search is to save your visitors’ time on typing. Smart Insights showed that more and more customers are using devices without a physical keyboard to shop online. These visitors will appreciate a search function that predicts the product or collection they are trying to find.
Advanced search & search results
While most companies will incorporate predictive search, not all are able to. So, if you fall into this category, you will need to make sure your site has a strong keyword search and well-structured results page that allows the customer to ‘drill down’ or filter their results further.
For the Graham & Brown site, we recommended that the customer be able to filter their search by any criteria, such as colour, designer, or room. This allows the user to perform a keyword search for ‘wallpaper’ and then refine their search to only display the specific styles, colours, brands or designers they are interested in.
Furthermore, we found that users often search for common product terms. These often exist as landing pages and provide a more useful experience, as discussed in part-one of this series
In these cases, rather than direct the user to the generic search results page, a better approach is to direct the user to the existing landing page that matches the customer’s search combination (e.g. black wallpaper). Now, users are shown a collection page connected with their search, but this now contains additional content and tips, not simply another screen of products. This is important as it drives search engine optimisation (SEO).
On the Graham & Brown site, we further improved and streamlined their search feature by introducing filters into the advanced search results and ensuring any search queries would display relevant landing pages to visit. After measuring the effect of these changes, we identified that: 1) after using the search, the time on site increased by 16.56% and 2) the number of pages visitors viewed after getting search results increased by 17.52%.
Make search about discovery
If ‘search’ is about exploration, it follows that you want to make it as easy as possible for customers to find and peruse the items they are most interested in, as quickly as possible.
The fashion retailer Zara takes predictive search even further and uses a feature we plan to include for Graham & Brown. When a customer searches for an item on Zara’s site, the predictive search takes over the whole screen with relevant products and related items appearing as the customer types. This immediately provides users with a wider range and more detailed view of products they are potentially interested in.
Using Apache Solr for predictive search
With Graham & Brown, our initial approach was to use Apache Solr.. This leading, open-source, search platform already provides auto-complete functionality However, we found that Solr provides incomplete suggestions for search terms.
As Graham & Brown is a multilingual website, we needed the predictive search to return results for the visitors in their preferred language and country. However, the Solr predictive search component only accepts a single search term in each request. It is not possible to pass Solr information about the language or country of the visitor, to apply in the search results.
Another constraint was that only a set of suggested search words were returned. For a customer typing, ‘red’, the only keyword suggestions would be ‘red paint, or ‘red wallpaper’; we wanted to include richer information in the results, such as price, product image and page URL. We also wanted to further combine this with details of offers, promotions and stock level.
How we overcame the limitations of Solr
To achieve a more comprehensive search function, and still use the Solr auto-suggest component, we would have had to create additional lookup requests to locate the missing data, but this would have an impact on site response time.
To present visitors with smarter, quicker and more powerful suggestions, we ended up creating our own predictive search ‘suggester’ component, running against the underlying Solr index. This returns a fuller set of search suggestions, with only one HTTP request, and takes approximately 50 milliseconds.
As a result, Graham & Brown customers now see results relevant to their corresponding culture and country, grouped by content type, i.e. products, collections, designers and articles. These results are also accompanied by some additional useful information, such as product price or product image, giving customers an improved search function that helps them reach relevant results.
Next: promotions, bundles & product management
Coming up soon. The next article focuses on how to incorporate promotions and product bundles. Elements that your marketing team will appreciate and be able to use effectively to upsell and cross-sell, leading to an increase in average order value.
