This guide is designed to help online retailers of every shape, size, and industry to increase their conversion rate and revenue.
It’s EVERYTHING we know about site search for ecommerce stores.
How much difference can site search really make?
To be blunt, your shoppers who search are simply more valuable than other types of shopper you have.
The average conversion rate of all shoppers is about 1.72%.
However, shoppers that interact with search at least once convert 3.5x – 4.5x more often.
To put it another way, ConversionXL found that site search in some cases made up nearly 40% of a company’s overall revenue while only being used by 10% of the site’s traffic.
ScreenPages did an independent conversion study and found similar results. Less than 10% of shoppers were using search and nearly 40% of overall revenue was coming from those visitors.
We’ve seen some stores with about 50% of their revenue being driven by through their site search.
Whether you’re at the low or high end – site search holds a tremendous amount of promise.
Is the shopping cart’s site search enough or do I need more? What’s the honest real-world value?
First, any site search is better than no site search. So, let’s start off with saying the site search you currently have through your ecommerce platform is better than nothing – it’s a good first start.
However, at the time of this writing, there’s not a single ecommerce platform that is offering an “intelligent” site search solution.
They all depend on varying degrees of basic keyword matching.
By comparison, speaking purely of our own customer’s results, over basic stock site search solutions, the difference is about 20% in conversion rates, and 15% in average order values.
An intelligent ecommerce site search solution leverages your user generated behavior with a host of other data points to influence how your site search behaves.
Essentially, we’re comparing a blind system vs. one that is continually optimizing.
A site search that never improves and continues to deliver the same results regardless of past performance vs. one that remembers and adapts.
Why Keyword Matching isn’t good enough
All ecommerce platforms offer site search that uses single keyword matching.
For example, when a shopper searches for ‘red nike running shoe’ the search bar will display all the results that have each of those keywords present.
That means, every product, regardless of what it is, if it has those keywords within it’s product data, will be displayed. The more keywords found in the data, the higher the position in the search results.
At first, that doesn’t sound so bad, but here’s the problem.
The search bar never recognizes the shopper’s intent of ‘shoe’, it’s only driven by keyword density.
So, instead of seeing shoes, which match the shopper’s intent, the shopper will see a mish-mash of shoes and other assorted items (gym bags, shirts, pants, shoe laces, etc).
Any item that contain these words in their searchable fields will display, regardless of what the product is.
If there is an product relevancy, it’s typically by chance.
Using Google Analytics, take your top few on-site search terms and test your store’s relevancy. Are shoppers seeing the products they expect or are there a lot of accessories muddying up the results?
How to deliver more relevant product results
First, let’s define product relevancy so we’re both talking from the same page here.
Simply put, relevancy is all about showing a shopper the products that best match their intent and expectations.
Obviously, easier said than done.
The end result sounds rather simplistic, but the methodologies and algorithms that get you there are far from it.
Product Awareness / Natural Language Processing Algorithms
The Product Awareness algorithm does two things really well.
- It reads product data in a your database and builds an understanding of what each item is.
- It translates each shopper query for it’s intent – deciphering what is intended to the product type vs a supporting attribute.
In other words, it can understand that a user searching for “red nike shoe” wants a shoe, and a user searching for “red nike shoe laces” wants an accessory for a nike shoes, and delivering the expected relevant results.
Intelligence / Machine Learning / AI Algorithms
Monitoring, learning, and leveraging the behavior of shoppers is another extremely accurate way to increase the relevancy of a given site.
While a product awareness engine is able to determine which products are relevant to a query, intelligence goes a step further and ranks products based on how shopper’s have engaged with them in the past.
An intelligence layer can leverage valuable data gathered by Google Analytics and other sources in order to “crowd-source” data.
Each time a visitor interacts with a site, searches, clicks a facet/filter, sees products, clicks products, adds products to cart, purchases, etc… data is gleaned that helps to rank products in future searches.
Types of search queries you’re getting
There’s a lot of nuance behind the psychology of shoppers who use search.
When you understand the “why” of the “how”, you can dramatically improve the shopping experience. It’s important to understand that the buyer’s journey can be anywhere from a few hours to a few months depending on a number of factors.
For example, people shopping for clothes will likely have a much shorter timeline from research to purchase than those shopping for laptops.
In this context, there are 3 basic phases of search visitors.
Exploration – Those in the “exploration” phase, are generally only window shopping.
As an example, a shopper on a fashion site might be looking for something “trendy”, but not know what to look for.
Research – Shoppers in the “research” phase are more specific.
For instance, they might be looking for a pair of shoes, but wouldn’t exactly know which color, brand, or style they want.
Purchase intent – These shoppers know exactly what they want to buy, and how much they are willing to spend.
For example, they might want a pair of white Vans for under $125.
The way shoppers search tells us a lot about which of these phases they are in.
Search query types = Feature searches, Broad spectrum, Use Case/Symptom, Precise
Broad Spectrum (Exploration or Research Phase) – Broad spectrum searches are often only one word in length.
A shopper wanting to buy a 4K laptop will search for “laptop”.
Feature (Research Phase) – Feature searches are used to find products that suit a particular need.
A shopper searching for a “4k touchscreen laptop” will do so if they don’t know the names of any laptops that have this feature.
Use Case/Symptom (Research Phase) – Symptom searches are very similar to feature searches. Rather than looking for a specific feature, shoppers are looking for help to solve a particular problem.
A search for “cough medicine”, is a good example.
Precise (Purchase Intent Phase) – Precise searches are for exact products or specifications. This shopper has already done their research, and is ready to make a purchase.
A good example would be a search for “512GB MacBook Pro 15””.
Should you Buy or Build?
Are 3rd party providers worth the money? I have a very smart dev team, I think we can build this functionality out in house, right?
If you have a technically proficient team, then site search might be something you’re thinking about tackling internally.
It’s possible, but before you jump down that rabbit hole, here are some things to ponder before you have your team flexing their skillset against SOLR or ElasticSearch.
- These search technologies are only a framework.
- There is no intelligence built in. ie. you’ll have to come up with your own algorithms. ( product awareness, nlp, machine learning, intelligence, synonym detection, fuzzy vs strict controls, etc…)
- What this means in practical terms, you’re starting from scratch and basic keyword matching – something your ecommerce platform is already offering – albeit without any intelligence.
- It’s never set and forget. You’ll have bug fixes. You’ll need to tweak, test, stress and modify your algorithms. And you can’t forget, shopper behaviors are seasonal, cyclical, and they flat out just change due to influences outside of your control.
- Then, unless you’ve also built out a user friendly GUI, you’ll always be dependent upon your dev staff for any site, data, or product changes.
- You’ll need to build out custom reporting tools to evaluate whether all that work is… working.
- And then there’s the additional hardware and infrastructure cost to host and manage the increased data load.
So while it’s certainly possible to build a solution, or to hire an agency to build one, the cost is incredibly higher than the initial appeal of open source software and the “benefits” that follow.
One last argument, you’d be stacking your dev team and your store against your competition who is leveraging a 3rd party team of professionals who have dedicated years to site search exclusively.
Search bar optimizations & best practices
Get more shoppers to use your searchbar, here’s how.
Actually Have A Search Field
Sounds silly, but topping our list of tips is to actually include a search field, and not hide it.
Hiding the search field is a big no-no.
Stores that only show the search icon or who hide their search bar in a flyout menu are lucky to see a 4.44% shopper engagement rate with search.
Simply adding the search field can triple your site search usage.
Only have ONE “search” input field
Get rid of any input field that is even remotely close in proximity to your search field.
There’s nothing more frustrating for a shopper than to enter a search and get the “this is not a valid email address” error feedback.
We recommend removing or moving any input field that isn’t the search bar out of view.
Use A Button
Preserving your design and only using an icon with your search field could leave you with a lowly 6.01% engagement rate.
Adding a button can boost that to 14.49%.
Even better, adding the obvious “search” text inside the button can give you a 16.17% engagement rate.
(Buttons that substituted text for a magnifying glass icon convinced 13.44% of visitors to search.)
White background and contrasting border
Make your search input field white with a high contrasting border color.
Doing so can take user engagement rates from a pitiful 8.78% to a whopping 18.82%
Include contextually relevant and helpful placeholder text
Including placeholder text inside the search input field will boost your engagement rates by 2% compared to leaving the search bar empty.
A 2% boost doesn’t sound like much, but it does add up, especially for such a simple fix. It’s a no brainer.
Pro tip: Make the placeholder text instructional and contextually helpful. For example, if you sell media, your placeholder text could say “search by title, publisher, artist, etc…”
Don’t cram it, add whitespace
Stores that give the search bar an appropriate amount of whitespace see engagement around 18.46% while stores who “try to save space” suffer with a 10.35% engagement rate at best.
A good rule of thumb, give the search bar enough room so that a finger could easily select the search input field, on desktop and mobile, without accidentally clicking on something else.
Center the search bar
The common industry convention is to place the search bar on the right side of header navigation. Surprisingly, this isn’t always the best option.
Putting the search bar front, top, and center can boost engagement rates to around 15.86% compared to the top right’s average rate of 13.43% or the top left’s 7.72% rate.
Keep the search bar in the same location
Keep it in the same location, on every page. We don’t have any fancy stats on this one, just some pure common sense.
Don’t make users hunt for the search bar when they engage with your store.
Adding an autocomplete that’s actually helpful
Autocomplete should provide two types of suggestions, related searches and product suggestions.
Vary Styling for Different Types of Suggestions
“Related searches” help the user complete their query more rapidly, and will take them to a search results page.
“Product suggestions” will take the user to a specific product page.
This distinction is important since it sets an expectation for the shopper. If they believe they’re completing a search, they will expect to see a list of products that they can sort and filter on the next page.
If the suggestions look identical, it will create confusion since many users will land on pages that they don’t expect.
We recommend that related searches be text, with a visual emphasis added to the characters the shopper has already typed.
Product suggestions should be placed below the related searches.
Add a border or a colored background to these suggestions for better separation. Also, include product images to convey that clicking will take the user to a product page.
Provide typeahead autocomplete suggestions and related search queries for your shoppers.
Shoppers aren’t great spellers. They also aren’t quite sure if they should use their lingo or yours.
Providing related and suggested searches while the user types will help eliminate misspellings, errant searches, and most of any confusion your shopper might have.
When you provide product suggestions directly in the autocomplete, you will see a considerable amount of your traffic navigating directly to those product pages and converting.
So much so, by providing product suggestions you can double engagement and conversion rates from autocomplete.
Limit suggestions so it’s manageable for the user (avoid long list)
Limit your suggestions to your top few and most relevant, anymore than that will burden the shopper and diminish their experience. If you’re suggesting dozens to hundreds of options then you’re not helping anyone.
Display pricing with product suggestions
Displaying the price with other product data helps shoppers understand what they are seeing is a specific product which will take them directly to that product page.
Keep the focus inside the search bar and not the first suggestion
Ensure that your search bar does not select or highlight the first autocomplete suggestion by default. By default, hitting the enter key should simply search for the query that the shopper has typed into the search field.
Remove Redundant Suggestions
You should do your best to avoid showing over granularization in your suggestions, especially if the suggestions deliver to the same results page.
Different spellings of the same word, plural vs. singular, or case sensitivity, bloated suggestions can start to populate for nearly any reason.
Get rid of them. Don’t confuse your shopper into thinking they’ll see different product results when they won’t, and don’t waste the digital real-estate on duplicates.
Support Persistent Queries
Persistent queries follow the user and retain the query in the search box after the shopper hits “search”.
This is important because the average search visitor enters more than 2 iterations of their search. Let them modify their search without needing to re-enter their entire search.
Make sure it doesn’t go offscreen
It’s easy to forget, but just because your website is responsive, that doesn’t mean your autocomplete will magically follow your CSS rules. Make sure your autocomplete dropdown doesn’t fly off or get partially hidden at any screen or window size.
Make It Accessible
Over 15% of the world’s population and approximately 19% of the US population live with a disability.
Regardless of how few people with disabilities you think might visit your store, ensuring your site search, autocomplete, and navigation is accessible with both the mouse and keyboard will only bring you positive results – not too mention a lot of good will from anyone who benefits from the added functionality.
Site search optimizations & best practices
The name of the game is delivering relevant search results. You can make every UI improvement above and use every tip below, but if your results aren’t relevant, there’s no saving that.
Here’s how to make your site search smarter and deliver better product results.
Customize What’s Searchable
Define what data fields your search bar is actually allowed to search. We recommend that most everyone have these fields set to be searchable:
- Name / Product title
- Sku / Product Code
- Brand / Manufacturer
- Category / Product type
- Any attributes relevant to your catalog: gender, size, color, make/model, author, artist, etc…
We generally try to avoid and make these following fields not searchable:
- Numeric fields with the exception of measurements
- Price – We of course leave this field filterable, but not searchable
- Description – This one tends to bring a lot of questions, but the description field is very problematic if it’s searchable. It injects a lot of noise that dilutes the relevancy of the product results. For example: A description field could have this line in it’s copy “This shirt is great for the beach, be sure to pair it with these pants” Now, this shirt has a high probability of showing with any “pant” related search. Of course there are exceptions, but 9x’s out of 10, we prefer to leave this field unsearchable.
Plan For Misspellings
Search bars don’t have spell checkers, but that shouldn’t mean the difference of seeing a hundred products vs. zero products.
The ugly truth is, the majority of shoppers will assume your store doesn’t have what they’re looking for before they assume they made a typo. It’s unfortunate, but it’s true.
They think it’s your fault for having a bad search, or that you don’t carry the product they’re looking for. You’ll get the blame before they even consider it was user error.
If your search bar isn’t optimized to handle simple misspellings, then you are losing a considerable amount of business from your engaged users.
Allow Product and SKU Code Searches
It’s a guarantee, you will have shoppers that try to search products by their sku or product code. If you’re not returning relevant results from sku searches, you will miss the sale and lose the shopper.
Bonus points if your search detects and serves relevant results for partial sku matches.
Create Synonyms and redirects for common zero results
Is “t shirt”, “t-shirt”, “tshirt” returning the same results or does your search bar get hung up with minor differences?
What about a bigger query difference?
What would your search bar deliver between “chesterfield” and “couch”? For example, do you carry “hoodies” or “hooded pullovers”? What about “chucks” or do you only carry “flat soled tennis shoes”?
There are queries your shoppers are using that are returning subpar to no results for products you absolutely do stock.
Dive into your search term report inside GA and know for sure what search terms your onsite shoppers are using.
Support content searches using redirects
You may be surprised to find that some of your most common searches aren’t for products at all, but for content pages.
Whether or not you have a blog, you certainly have pages that don’t have any products. Some of the most common content searches are for “return policy”, “shipping policy”, and “contact us”.
Set up redirects to handle these searches. This alone will help you support the majority of your content searches.
Account For Singular & Plural keywords
Should “book” and “books” or “dress” and “dresses” be handled the same or should they display a different set of product results?
We see good reason to treat singular vs plural as the same until there is a specific use case or reason to do otherwise.
Account For Special Characters ( like measurements )
Will your search return the same results for “3 foot cable”, “3’ cable”, and “three foot cable” ?
Setting up synonyms and redirects for every special character variation across all numerical values is virtually impossible.
Ensure your search is smart enough to recognize and account for all these variations.
Avoid Dead Ends
Sometimes shopper’s just search for the wrong things because they don’t know better.
Suggest a different route the shopper can take.
- Provide “Did you mean” suggestions for semi relevant and related searches.
- Highlight your popular categories and navigation for browsing.
- If the shoppers are searching for common discontinued products or items that you do not carry, create a custom landing page with a relevant alternative.
- Provide an easy method of contact for shoppers to ask questions directly – including a live chat feature is extremely helpful in these scenarios.
You might have to get a little creative depending on your use case, but delivering a shopper to a dead end should be avoided at all cost.
When Dead Ends are inevitable – confirm their search.
Often times, people land on these pages due to incorrect spelling or typographical errors. Displaying their search in large, bold font near the top of the page gives the user an opportunity to see the search they entered and realize their mistake.
Even when the user did not make an error, knowing what they searched for will prevent them from entering the same exact search a second time. This helps them to understand that they need to make a modification to their search in order to find the products they want.
Make It Fast
Reduce the overall size of your page and make the content your search is delivering as efficient as possible.
- Use a fast hosting service.
- Avoid redirects when possible.
- Minify and concatify your script files to reduce http request.
- Optimize your shopping cart platform’s configuration.
- Specify your image dimensions. Optimize your images sizes. Strip out unnecessary image metadata.
- Use a CDN for your images
- Enable caching
- Compress your data with gzip or other compression methods
Learn from your ecommerce site search
Trends, products, and consumers all change rapidly.
Begin checking your data and creating reports on a regular basis to identify your wins and losses – ie. optimization opportunities.
Your shoppers are giving you a goldmine
Every shopper that visits your website leaves data behind. This data should be consistently used to improve and polish relevancy via site search.
Items viewed, purchased, searched, filters selected in the navigation, discount vs. new product sales performance – it’s all there. Start using it!
Monitor Zero Results
This is a must. You will have shoppers that sneak past all your defenses and they will create zero result searches. It’s okay. It’s not the end of the world, but you should never let it persist.
You can minimize zero results by:
- Ensure your site search is smart enough to deliver “Did you mean” suggestions for common misspellings or partial searches.
- When “Did you mean” fails, you can use synonyms. For example, “Did you mean” might not pick up that a “chesterfield” is actually a search for a “couch.” Watch your zero results reports, you’ll find queries that you can easily fix with a synonym.
- Sometimes, adding a synonym won’t be good enough. Even if the results are relevant, without including the shopper’s keyword query within the results, there can be a disconnect. Your goal is to shrink the gap between what a shopper is expecting and the products being displayed. When it makes sense, it can be beneficial to add the zero results query language into your current product labeling.
- The odds are good you already have the shopper’s language within your product data. The problem is, it’s not being utilized because those field values are not searchable. Modifying searchable fields can help minimize the zero results being displayed.
- You will have times when the shopper’s query is nowhere in your data, but it should be. Creating and using the custom fields that your platform makes available is a good way to enhance the product data to capture these types of shopper queries.
- Okay, you’ve successfully eliminated the zero results, but what’s being displayed isn’t quite right. When a shopper’s query has a specific expectation, creating a custom landing page with optimized product results could be the key to turning your shopper into a customer. After all, you’re doing the work to eliminate zero results. Go the extra step. Optimize what products are being returned for the queries that make sense to do so.
- Finally, there will be searches for products you don’t carry, but maybe you should. If you have a high volume of searches for a product that isn’t in your catalog, but would make sense to offer, you know you already have the market to begin selling it. Zero results could reveal the low hanging fruit opportunities that shoppers are already looking at you to fulfill.
The list of things you can do when monitoring your zero result searches is hundreds of items long. You can do whatever you like, but the point is, do something. Don’t let the same zero results continue to happen.
Review popular searches with low conversion rates
This is a huge opportunity to increase conversion rates that many retailers are missing out on.
As you review search queries in Google Analytics, take a look at searches that have very low conversion rates, and especially those with zero conversions. Likely, there’s a problem with the results that are loading, even if it’s not resulting in a “no results” page loading.
Creating redirects for synonyms with better results or custom-merchandising a landing page are great ways to increase conversions on these pages.
Improve your data feed structure
Every shopping cart has a data feed that powers the product pages and contains all the product data needed to display items. If this feed is rudimentary or missing data, the storefront will reflect that.
Improving the data feed can in conjunction improve many things, like product search relevancy, category navigation, product imagery, product descriptions, SEO, merchandising, etc. The data feed is the core of the shopping cart, neglecting it is a crippler.
Pro tip: Normalize your data
Irregular capitalization and misspellings inside your data can be very aggravating to shoppers.
For example, options of “blue”, “Blue”, “BLUE”, “bleu” should all be relabeled as “Blue”.
Normalize your data so that your product display is clean and clear. Once you’ve chosen a standard practice for how you’re entering your data, stick to it.
How to find your top searches
Your search terms report can be found in Google Analytics under Behavior > Site Search > Search Terms.
If you’re not seeing any search terms reported, make sure Site Search has been properly enabled within your Google Analytics account and the necessary changes have been made to the tracking code specific to the site search solution currently in use:
From this view, you can see how many times visitors searched for each term, and what they did after completing these searches.
You need to review and type these search terms into your search bar. Are the results relevant?
Next, it’s important to know the bigger picture surrounding the behavior of your search visitors.
In Google Analytics, navigate to Behavior > Site Search > Usage. Here you’ll see a comparison of search visitors vs. non-search visitors.
You’ll be able to see what percentage of your visitors are using search, what their bounce rate is, how much revenue they’re generating, and what percentage of them are converting.
Are these numbers generally higher than non-search visitors?
In most cases, search visitors will have much lower bounce rates, and much higher conversion rates. On average, search visitors convert at a rate around 3.5x higher than non-search visitors. Search visitors only bounce about 10% as frequently, indicating they are much more engaged.
Are you below those benchmarks? Focus on fixing search relevancy.
Are you hitting those numbers? Fantastic, now, increase engagement and get even more!
Average 4.6 Rating on G2 Crowd