An article by Darrell Wilkins on 12 Jan 2022
Summary: Search is a desirable feature for any large website. But it’s hard to do well. Don't use search as a substitute for good navigation. And if you are considering adding search, invest in doing it properly.
Search is an essential feature for any large website. If you have over 200 pages, you should consider adding it. Around half of all people are search dominant, meaning it is their preferred way of finding what they are looking for.
Search is not a panacea, however. And you shouldn't implement it at the expense of good site navigation and way-finding. If your search experience is poor, your prospects can't quickly find what they need and they will leave.
Does Your Site Need Navigation or Search?
Websites that cater to their user's preferences achieve better results. So, should you prioritise site navigation or search? Studies show that 43% of users go immediately to the search bar when available. And the success of Google, YouTube, and Amazon (the world's top 3 search engines) show how important it is. As such, it's reasonable to ask if you should care about navigation if you also offer search.
The answer is yes. You absolutely should care about navigation. Remember that around half of your audience prefer browsing over searching. In addition, there are three other reasons navigation is essential:
1. Way-finding
Good navigation allows users to move around your site and gives people a sense of where they are and what they can do next. This is a critical feature of good user experience design. And useful even if someone has used search to find a particular page.
2. Knowing what to search for
People are not very good at formulating effective search queries. It's not uncommon for people to browse the navigation to see the terms you use and then use what they find to formulate search queries. Navigation can aid your users’ understanding of the search space, making searches more successful.
3. Searching is harder than browsing.
Search requires you to recall information from memory to formulate each query. Navigation replaces recall with recognition — which has a much smaller cognitive load. Not only that, search requires you to move your hands from mouse to keyboard and then type. While that may not sound like a big ask, typing is error-prone, especially on mobile, but it's more difficult than navigating with a click. Even if you're an expert.
Before you implement search, design an excellent navigation system. Then, not only will you serve the people who prefer to browse, but you will also improve the search experience.
What Does Good Site Search Look Like?
Before we examine what good search looks like, it's worth understanding what search is.
When users enter a search query, they have a specific goal in mind. They want to find, learn or do something. A search engine must understand what they intend by their search term and match it with relevant pages on your site.
We’ll cover intent more later, but let's look at matching, relevance and search user interface (UI). All three contribute to better search functionality.
Matching
Simple search is easy to do. This is where you directly match the words typed into a search box with words that appear on a page. Even mobile phones are powerful enough to search a large document and instantly match text strings.
This is what most out-of-the-box or plug-in search systems do. If you don't have many pages on your site, this might be good enough as you are effectively just filtering through results.
If you have hundreds or thousands of pages, however, just matching text strings does nothing to help you present relevant results. And delivering relevant results is the primary job of a search engine.
Relevance
Eye-tracking research into search engine results pages (SERPs) has shown that:
- In 20% of searches, participants only looked at the first result.
- 41% looked at results 2 → 6
- Only 13% looked at results 7 → 11
- 2% looked at results further down the page.
These stats only refer to what people looked at, not what they clicked on. So if you don't provide relevant pages near the top of your search results, they are pretty much invisible.
How does your search engine decide what is relevant? How do you rank your pages so the ones most relevant to your user's goal appear at the top? Basic text string matching doesn't help. To display the most relevant SERPs, you’ll need a more sophisticated search engine to assign priorities to different types of data.
For instance, you might rank a page higher if the search term appears in its title or summary. In addition, you could rank a page based on the number of times the search term appears on the page or based on the date you publish the page.
These techniques can certainly improve your search results. However, they are based on many assumptions, such as a newer page being more relevant, which might not be the case. To rank things well, you need to understand the intent of the searcher.
For now, it's enough to say good search presents relevant pages that match users intent. But that is only half the story. The second part of great search is the interface you provide.
Designing the search UI
Compared to understanding intent and delivering relevant results, designing a good search UI is pretty simple. There is also a lot of well-tested best practice to lean on.
- Search should be available from every page.
- The best location for your search box is the top right of the page.
- Use a search box with an empty type-in field, not a link or an expandable icon.
- Use a button that says search, not a magnifying glass.
- Make the search box wide enough for your typical queries, so the text doesn't have to scroll.
- Unless you have vast numbers of pages (hundreds of thousands), you should search the entire site rather than letting users limit the search to results from specific areas of the site.
- Provide filters on your results pages to refine the search further.
Even though Search UI design is a well-understood area, getting itI right is essential to the overall experience.
So now we know what good search looks like and can see it's pretty straightforward to implement the basics. So, why is it hard to do search well? It all comes back to intent.
Why is Search Hard to Implement?
To serve relevant search results, you must understand users' goals when they enter a search term. Are they looking to:
- Do something — look for a job or contact a partner.
- Learn something — learn about a specific feature or how you might help them.
- Find something — Find a specific page or diagram.
We have to understand what our user’s mean to search for while accounting for the peculiarities of human language.
Understanding search intent
When you have an idea of your user's goal, you can serve more relevant results. And as we know, relevant results are everything. Take the word “Java” for example. Java can be:
- A programming language
- An island in Indonesia
- Coffee
A very basic search engine could match the search query Java with all the pages on your site that contain the word Java. But if you returned me pages about programming when I was looking for information about an island paradise, it wouldn't just be irrelevant — I'd be miffed.
Now, if you add visit, write or drink to your Java query, it's much clearer (to a human) what you mean.
- Visit Java — I want pages about the Island.
- Write Java — I want pages about the programming language.
- Drink Java— I want pages about coffee.
Humans can associate and infer context well by that extra word. We are natural pattern matchers, after all. Computers (generally) are not. They work based on rules. And considering there are more combinations of 100 words than there are atoms in the universe, you can see what a huge task they have.
Here is another example. If you perform a search on your site for a whitepaper, are you looking for a whitepaper that the company has written or an article referencing someone's whitepaper? The same search query without context can mean very different things.
How our strange language impacts search results
Understanding intent is not the only hurdle to overcome. Our language is rich and complex and has all sorts of weird and wonderful things that can trip up search engines.
Homonyms
We've touched on homonyms (words with the same spelling but different meanings) with our Java example above. Some other examples:
- Bright — Very smart or intelligent — or — Filled with light.
- Express — A fast version of something — or — To show your thoughts by using words.
- Well — In good health — or — A source for water in the ground.
Matching those words to a page on your site is easy. However, understanding what the user meant when they used them and serving relevant pages is not as straightforward.
Synonyms
Synonyms are a word or phrase that means exactly or nearly the same as another word or phrase. For example, shut is a synonym of close. If I searched for shut down, a simple text matching search would not return a page with closed, even though it means the same thing. That is not a great experience.
Like synonyms, people often use abbreviations. For example, your page might mention Chief Executive Officer, but your user entered CEO as a search query. They mean the same thing, but a search engine without synonyms support would see them as entirely different.
Capitonyms
Capitonyms are words with different meanings if their first letter is capitalised.
Consider the following:
Before we started our march in Turkey last March, we ate turkey.
You and I can parse that sentence quickly, but constructing a set of rules for a computer to understand that phrase isn't. Granted, Capitonyms are an edge case, but a good illustration of the complexity of our language and why search is hard to get right.
Spelling
Some words are hard to spell. Some are spelt differently depending on where you come from, like colour and color. Sometimes you're not sure if there is a space in a compound word, as in 'white paper' or 'whitepaper'. Sometimes people make typing errors or spelling mistakes. Simple search can't deal with these, which diminishes the user experience.
It quickly becomes apparent why search is so hard to do well. Google has spent billions of dollars on it, and while it’s often excellent, sometimes you still can't find what you are looking for. It's not all bad news, though.
What Is Involved In Making Search Work Well?
We've already talked about what good search looks like. The UI is relatively straightforward. The only tricky part is getting your search to provide relevant results. To achieve the best search experience, here are three starting points.
1. Use good technology
Attempting to write a search engine from scratch would be insane. But luckily, you don't have to because some brilliant people have worked on solutions for years. In addition, there are various open-source solutions and compelling paid offerings you can implement.
These help with the two most challenging parts of implementing search across your site: Understanding what someone was looking for and delivering relevant results. Some of this is done with complex algorithms, but increasingly machine learning is playing a big part.
But while technology helps immensely, it won't solve all your problems.
2. Preparing your content
Companies put a lot of effort into Search Engine Optimisation (SEO), preparing their content and optimising keywords to rank well on Google. You have to do the same thing with your on-site search engine.
This means reviewing search logs, understanding what your users are searching for and manually tweaking your search engine to affect how pages rank. You'll also have to look for the synonyms, acronyms, and spelling mistakes your users make. Most sophisticated search systems will allow you to map these to the words or phrases you're actually using.
If you do all this, then, over time, you can produce highly relevant search results. You'll probably learn quite a lot about your prospects and customers, too.
3. Keep at it
Effective search is not a set-it-and-forget-it system. You have to keep at it for as long as you want your search to work well. You should constantly tweak your search index, ranking algorithms and content to ensure it is working well. It would help to extend your regular website user testing to cover the search feature.
Should You Implement Search On Your Website?
If your site has 200 pages or more (including your blog), then providing you invest the time and resources, yes, you probably should. Don't underestimate the effort involved in creating a great search experience, but your customers and prospects will reward you with their business if you do it well.