“Not all who wander are lost.”
Good stuff for bumper stickers or if you’re trying to write a generation-defining bestseller and get Reese Witherspoon to play you on the big screen. In ecommerce, however …
A wandering customer only means one thing: a lost customer.
Having been conditioned by Google, today’s consumers have high expectations when it comes to search functionality. It’s far from surprising then that — as Nielsen Norman Group’s exhaustive E-Commerce User Experience found — “Most e-commerce customers go directly to a site’s search tool to find products.”
Given that none of the leading ecommerce platforms have strong native search, this piece focuses on what to look for in a third-party search solution along with core …
Ecommerce Site Search Best Practices
- Natural Language Processing for Better Results
- Merchandising for Products, Attributes, and Images
- Machine Learning for Personalization
- Predictive Autocomplete for Faster Results
- Strong Error Tolerance for a Smooth Journey
- Comprehensive Analytics and Reporting for Optimization
- Simple Integration for Faster Time to Market
- Mobile Search to Turn Browsers into Buyers
Keep reading learn more
But, if you’d like to go behind the scenes and find out how onsite search is just one part of what helps the average Shopify Plus merchant grow between 126% and 274% YoY …
Then download the full recordings and slide decks from our two-part webinar event, Growing Your Entire Online Funnel.
What Is an Ecommerce Site Search Solution?
An ecommerce site search solution is a third-party tool used onsite to sync customer search queries — e.g., natural language — with product titles, variants, descriptions, images, videos, SKUs, and reordering codes.
Following in the footsteps of other ecommerce technologies, onsite search has become mainstream in recent years. There is now a host of options available for merchants when previously there were just one or two main providers that were cost prohibitive and required complex, resource-heavy integration.
Today, options exist for merchants of all sizes, with even enterprise-level solutions available from as little as ~$200 per month. This availability has driven a surge in popularity and usage.
At the forefront of this ‘disruption’ at the mid-level and enterprise-level ends of the market are solutions like Klevu — where I serve as a Solution Architect — and Algolia. Legacy providers have become less relevant, due to not being as agile (in terms of releasing new functionality) and pricing.
Algolia is a good example of a wider search technology, which provides excellent indexing capabilities and speed to all. Klevu is more ecommerce-specific and provides exceptional accuracy, primarily through the use of natural language processing.
Why Does Site Search Matter in Ecommerce?
Onsite search represents a strong opportunity for retailers going into 2018. The rise of mobile commerce demands faster and easier routes for finding things. Even on desktop, search drives higher conversion rates.
Various reports on the value of search for merchants bear this out.
- On average, users who complete a search are 1.8x more likely to convert
- Site search visitors can generate as much as 13.8% of overall revenue
- In the case of detail-oriented products, while less than 10% of visitors may perform onsite searches, upwards of 40% of a site’s revenue came from them
From my own experience with sites in the enterprise space, I’ve found the average conversion rate increases to be closer to 3.5x non-search visitors, with around 5-10% of visitors using search.
In other words, retailers have an opportunity to drive more revenue through search, especially for those who haven’t spent time optimising this area. And, at the opposite end of the spectrum, there’s no better way to frustrate and disappoint customer than serving up irrelevant results or even 0 results for relatively straightforward queries.
Best Practices for an Ecommerce Site Search Solution
1. Natural Language Processing for Better Results
The demand for natural language processing (NLP) within search has increased considerably recently — allowing for accurate results even when the user doesn’t really know how to describe what they’re looking for. Having long been adopted by the likes of Google for its organic search, NLP-based search has now made the transition to ecommerce, driving real change in this often overlooked part of the online shopping experience.
NLP algorithms are based on context and relevance, rather than simply on the presence or absence of keywords in product names or descriptions. In simple terms, that means that an NLP will be able to extract meaning from the query to understand that a visitor who types in ‘red jumper’ is happy to look at sweaters and pullovers, even though the query doesn’t contain the term ‘jumper’.
The same applies to color and being able to understand variations of red. This can be really valuable for retailers, particularly with more complicated queries — an example can be seen below from a Klevu demo store.
This example shows how NLP is being used to understand more about the query - this example also benefits from more product catalog data being indexed (such as product reviews, pricing, all other available descriptive attributes etc).
Here’s another example from the same store, which is from a proof of concept implementation with an IR100 retailer.
2. Merchandising Capabilities for Products, Attributes, and Images
Often one of the main issues that leads to a third party search solution being used is the lack of merchandising options for results, with no way to change the ordering of products being returned and a lack of information being presented generally.
Newer third-party tools address these areas and have built-in features to allow retailers to ‘searchandise’ their results properly and weight different items and specific attributes.
Klevu, for example, allows users to assign ‘hero SKUs’ at query level, boost products based on specific rules (based on tags and meta fields) and also assign weightings for specific queries (e.g. boost this red Nike t-shirt across all relevant queries). Another merchandising feature that I think is important is the ability to add suggested listings to error results.
Comprehensive options around filtering, the displaying of product labels and handling of things like variant information are also handled well by the top tier of providers.
The features that are important in this area include the ability to:
- Assign hero SKUs across specific queries and groups of queries
- Visual merchandise specific queries and groups of queries
- Create rules for boosting a specific selection of products (across individual queries and groups of queries)
- Handle variant-specific queries (e.g. “XS Nike training t-shirt” or “black Adidas Climacool jacket”
- Include product labels (ideally that have been assigned within Shopify Plus)
- Include attribute-based (tags and meta fields) filtering in the quick-search interface and on the results page
- Add recommended results to 0-result-search errors (ideally personalized)
The payoff for this level of attention to detail can be handsome.
International Military Antiques, for instance, credits InstantSearch+ — who helped them create detailed search rules for over 7,500 products with unique and often overlapping attributes such as type, time period, and nationality — with a 600% lift in conversions among visitors who use their onsite search:
3. Machine Learning for Personalization
Most enterprise-level retailers have already adopted machine learning-based solutions at some level, be it via onsite product recommendations, category merchandising, or onsite messaging and content.
Search and product recommendations are probably the two mainstream usages.
I tend to work very closely with NOSTO, who is a big player in the personalization space and they were one of the early third parties to really promote machine learning.
Personalization is a new area that’s been introduced by a number of the different providers, which essentially promotes products based on 1:1 user behavior. So, one example could be, if I’ve been interacting with men’s Nike products, the associated products would then be boosted for other queries. I think this will become a mainstream feature in search in the coming months and it’s another good, value-adding feature.
The use of machine learning adds a second layer of accuracy, prioritizing products based on their performance and also ensuring that results are improved over time, based on the ‘learning’ from user behavior (e.g. the products that are being clicked or purchased most frequently).
Most third-party search tools also support manual ordering logic and merchandising, to allow for other factors (such as inventory and seasonal popularity).
4. Predictive Autocomplete for Faster Results
This example shows a rich instant search interface which allows for category results, query suggestions and also content results (not visible for this query). You can also see that product reviews are being displayed.
These suggestions can kick in after just two or three characters have been entered, bringing the most likely results right to the customer’s immediate attention, with various product attributes such as thumbnails, various naming fields, configurable options (such as colour and size), product labels, product reviews and even add-to-cart buttons (as can be seen in the Micro Scooters example below).
Suggested categories can also be displayed in the quick search interface, as well as non-product search results, such as customer service pages, blog posts, size guides or even recipe ideas. Some search tools can even show the full result set right there in the dropdown, with scrolling and faceted search filters also available within the dropdown — as can be seen in the example above.
This example on Soho Girl provides a full-page experience on mobile devices, which represents a really positive user experience that can significantly speed up the process of getting through to the product detail page, particularly against conventional category-led journeys.
5. Strong Error Tolerance for an Uninterrupted User Journey
Being able to handle errors without manual interception is another key requirement for a third party solution, which is often aided by the NLP and machine learning features outlined earlier.
It’s also important to have reporting around error keywords and the ability to assign synonyms (or automatically switch keywords).
6. Comprehensive Analytics and Reporting for Optimization
As with any core premium solution, it’s important that to understand the value they’re providing, as well as the general performance of the function. Generally, most of the SaaS solutions provide strong reporting, here are some of the reports I’d suggest looking for:
- Query usage and search term counts
- Filtering usage and number of filters used
- Top performing queries (conversions)
- Top performing products (conversions)
- Top performing products (clicks)
- Under-performing queries (conversions)
- Where queries are being performed (pages)
- Where queries are being performed (geographically)
- 0 result queries
For example, high volume search phrases that convert weakly are worthy of attention, as they could indicate a product availability problem, or they could lead to new buying decisions for future product expansion.
These can also dictate merchandising decisions and actions. Looking at where searches are being performed from is another good report to look at (also available via Google Analytics), as it usually highlights an issue with product or merchandising.
These are just a few of examples of the kinds of insights that can be gleaned through mining the often available data within third-party search providers. With appropriate resources to regularly analyze and act upon the information pulled from onsite search, significant, measurable gains can be made.
7. Simple Integration for Faster Time to Market
For a new search solution to deliver value quickly and cost-effectively, it needs to be simple to integrate into the ecommerce platform and to configure.
A lot of the more well-known third-party solutions have integrations with Shopify Plus and the only custom development work is around the search results page template, the quick search interface etc.
Klevu (who I have the most experience with), provide all of the functionality I’ve discussed via their app and are due to release a solution for category merchandising (for product list pages) and different template options for the results pages.
There are some solutions out there that have far more complex integrations, but my experience is that these are only really worth it for huge stores, and even then, often the more straightforward technologies offer the same capabilities.
8. Mobile Search to Turn Browsers into Buyers
More people are now accessing the internet through smartphones and mobile devices than through desktop computers. As such ecommerce is critically dependent on a robust mobile strategy.
I’ve worked on a number of stores recently where we’ve moved to using search as the primary user journey on mobile, with traditional menu-based navigation being harder and slower to use in a mobile environment.
If you’re able to make use of the rapid predictive capabilities already outlined and also suggest the use of more detailed queries (such as a specific brand or style) search should provide a far quicker and cleaner journey on mobile.
One example in particular really benefited from this move and they’re able to attribute a considerable uplift in conversion rates on mobile to increasing search usage.
The retailer already found that users who completed a search were around 900% more likely to convert and they were able to drive more revenue by increasing usage (on both desktop and mobile, but mobile being the more impressive), be it at a lower than 900% rate.
Retailers like Amazon and eBay have set a precedent around search being the primary user journey across the site, but more and more large retailers are going this way on mobile.
Obviously, for this to be a good option, it’s important to ensure that your search function is capable of processing more complex queries and you’re able to merchandise results effectively.
Ecommerce Search Solutions
For larger retailers, these best practices are likely to be required to get the most out of the search function, which also requires input around UX and merchandising. The NLP and machine learning components, in particular, can reduce the amount of manual work required and drive revenue by processing complex queries.
With so many third-party providers vying for business in the enterprise market, here are some of the better-known specialist third party solutions that I’d recommend for Shopify Plus merchants in particular:
- Klevu (Shopify Plus Technology Partner)
- InstantSearch+ (Shopify Plus Technology Partner)
- Findify (Shopify Plus Technology Partner)
- Celebros (Shopify Plus Technology Partner)
- Nextopia (Shopify Plus Technology Partner)
- Algolia (Shopify App with 5 Star Rating)
About the Author
Paul Rogers is an ecommerce Consultant and the Founder of Vervaunt, a small London-based ecommerce and paid media agency, as well as a Solution Architect at Klevu. Paul works a lot with Shopify Plus and recently wrote this piece, which outlines some of what he believes to be the core benefits of using the platform.
- A Crisis Communications Plan for Brands
- Ecommerce Agency Automation: Saving Time, Selling More & Launching Faster
- The Seven Signs Your Supply Chain Is Hemorrhaging Money & How to Fix It
- Direct to Consumer vs Wholesale: Customer Experience Over Competition
- Hypefest 2018: O2O Examples from the Forefront of Marketing & Retail
- 5 Easy Tips For Getting Started With Conversion Rate Optimization
- Wholesale Ecommerce: What is It and How to Start?
- How to Avoid the Hidden Cost of Black Friday, Cyber Monday Sales and Increase Customer Lifetime Value through Personalized Email
- ‘Emotional Ecommerce’ Increases Repeat Purchases 80%, Retention 58%
- 10 Best Omni-Channel Retailers and What You Can Learn From Them