Designing systems that work.
Contact us to get started today.

Top 5 Mistakes in Search Tool Design

Lot’s of organizations deploy search to help their clients or staff members’ discover and navigate their content. By understanding the common mistakes outlined here, and follow these suggestions to fix them, hopefully you will be able to make improvements to the search tools your company deploys. Whether the search is for a web site, an entire intranet, a document or knowledge management system, or just a simple help tool, there’s several common mistakes that appear again and again in the design of the tools and their user interfaces. 

These problems appear in search tools, sometimes regardless of the search engines deployed behind them. Even if you’re using a powerful and capable enterprise engine like Autonomy’s IDOL or the Endeca Information Access Platform, or just a simple keyword tool like Apache’s Lucene, these mistakes are choices that are made when designing and building the a search tool and it’s user interface that can be avoided.

Most of these are written assuming some kind of internet facing search tool for a web site or product site is being offered, but many of the problems (and ideas for solving them) apply to other search tool scenarios as well.

1.      Zero results found

search-tool-mistakes-zero-results-found

 

 

 

My least favourite screen from a search tool. What do you mean? Nothing found at all? Why ever not? Surely there’s something to offer. Just because the search engine didn’t return any matches, that’s never an excuse to display a screen like this.

First of all, always check your search engine is working well – there may be reasons more people than normal are seeing this screen. It’s always worth performing periodic acceptance tests on the site and the search engine itself – they tend to get neglected once launched and if the engine has stopped working properly, you may not realise. Make sure your search engine supports synonyms and has all of your products and services entered into the engine so it’s capable of finding relevant pages. Make sure all your content is indexed properly – should the search have no result, or are you missing pages?

But even if you’ve done all this, there’s still no reason user’s have to hit the dreaded zero results screen, even when your search engine comes up blank.

There are a number of things that can be done to mitigate this experience.

One is to back off from ‘all’ words to ‘any’ automatically (if your search engine is so picky about such things) – be sure to explain to your user’s that you couldn’t find anything with all of their terms in, but here’s some that might help or be close.

Offer ‘advertised’ links. These are results that trigger on certain keywords anywhere in a search string and show a search result pre-defined by someone in your business (which usually includes a link and some descriptive text). By showing things that might be useful specifically related to some of the searcher’s keywords, you might be hitting on exactly what they’re looking for. It’s worth making it clear that there’s a difference from these results vs. the normal ones from the search engine.

Another approach is to offer up top results from previous searches. You are tracking all your popular searches, right? Just by listing the things users most commonly look for on your site might help out – it’s certainly better to offer these links than nothing at all. If worst comes to worst, a basic site map might help.

Don’t create fake results, make it clear to your user that you couldn’t find a match, but perhaps these popular links would help.

Zero results leaves a user at the end of a road with no where left to go. At the end of the day, you want them to find what they’re looking for and you want to give them options and tools to try and find it, even when your default search engine can’t help.

2.      Searching only matches content – what am I searching?

The problem I’m referring to here is that issue where you’re searching for a form, tool, or section on your website, but the search engine is only looking for keyword matches in your web page content. As such, a search for ‘contact form’, or ‘shopping cart’ might return no results at all.

This is perhaps more of an engine issue than a user interface problem – but this issue occurs in part because it’s not clear what the search tool is searching for me, vs. what I think it might be. What did I just search for?

If you make it clear to your users that this tool is only a product search (for example), I won’t use it to try and find sections of your site. Or, on the flip side, if it’s only searching your help pages and marketing content then make it clear this won’t find products or won’t find tools on your site.

These issues are easily mitigated with this kind of communication to the user, and this can be worked in easily by labelling your search box and search results clearly ‘Search our products:’, ‘We found 5 products that match your search:’, and so on.

An even better way to solve the problem is to have a comprehensive search tool. If you’ve got tools on your site, like a shopping cart, location finder, contact forms, or whatever it might be, make sure that your search box lets your users find them. By simply including a number of web pages in your content that describe the tools well, and have links to them, at least the user finds the web page that gets them to where they want to go. Make the title of the web page the same or similar to the target tool so that the result in the search list shows up as ‘Contact Us Form’ rather than ‘About the Contact Us Form’ or something equally misleading.

If you have a tool only offering product search, make sure your results page has clear navigation options available to get to tools that might be needed on your site – shopping cart, check out, wishlist, find a store, or whatever they might be.

Again, ‘advertised’ links can help here. If you don’t want to create content pages, just make your search engine shows a special link to the right tool page in your web site when the right keywords are spotted. 

3.      Unusable or bad search results layouts

There are many bad ways to layout results in a search interface. It’s more common to find many things wrong with the layout of a search results screen than to find just one or two. Like any user interface, some of the basic principals of user interface design need to be applied!

Here are some problems that I’ve seen search results pages suffer from, and this is far from a complete list:

  • Excessive padding of white space around each result (white space is vital, but if a list of 5 results is spread out too much, people might not even scroll to look through them).
  • Too few results displayed – if you do have space left, then show some results, excessive pagination is just annoying (most screens are bigger than 640×480 now, you know!).
  • Giant product or results pictures next to each result, or pictures you just can’t make out at all  – the idea is to get across what the product is, not to see it in all it’s glory or to require a magnifying glass. Worse is when the picture shown hardly means anything at all.
  • Search tool mistakes: Meaningless picturesDivider bars between each result that aren’t even needed to visually separate the results (usually the right amount of white space or some kind of bullet works fine – the bullet might even be the relevance ranking indicator).
  • Huge text summaries that go far beyond the call of duty in terms of giving enough context about the document.
  • Text summaries of varying sizes – if I’m visually parsing a list I need to clearly see where each item breaks. Each item in the list needs to be the same size. Having summaries or pictures that vary significantly in size mean that the items are not consistently laid out and it’s hard to see as a list. Truncate summaries, resize product pictures. Be consistent for each result presented.
  • All the result information, including the summary and other meta data, put in the same size and colour font so that it’s hard to distinguish between the types of information presented and so that results take up more space than they need to.
  • Titles shown all starting with the same text, so distinguishing results is nearly impossible – this is a meta cleanliness issue, but last minute parsing and cleanup can go into the UI if there’s no other choice.search-tool-mistakes-poor-navigation-of-results1
  • Bad control layouts that put infrequently used controls front and center, navigation that interrupts getting to the search results, and similar weak user interface design issues.
  •  search-tool-mistakes-poor-search-tipsUseless or irrelevant search tips or advice. I often see a search tips box that contains things users can try just because the search engine is so bad – and they’re shown regardless of what’s happening on the screen. I like these tips here – I always expect my customer to know exactly what I mean by ‘try a synonym’. These tips appear regardless of the number of results (my result might well be right at the top). In this case, these just distract the user from finding the information they need. A better approach is to offer a link to an advice or help popup, such as ‘Can’t find the what you’re looking for?’.
  • Never ending lists – whether it’s groups of results or the results themselves, just putting a massive incoherent and unreadable list on the screen isn’t helpful. Divide up results, make it clear which ones are related if needs be. 

Some of these issues can be summed up neatly as ‘bad use of real estate’. There’s only so much screen to go around, and there’s a lot to present. At least, we hope, several matching search results and all of the associated meta information (see below). In addition, there’s guided navigation options (see below), important links (see above), and branding and navigation for the rest of the site, of course.

On the flip side, cramming things in for no reason, shortening information to the point of being useless, or assuming everyone has tiny screens is just as bad.

I always am sad to see a search results layout that doesn’t make best use of the space available – it just makes the process of finding the information a user is looking for harder and more unwieldy.

If you feel the need to make results stand out by showing larger pictures or more data about results, then perhaps do that kind of ‘profiling’ for just the first three results – or even just the top result.

Bear in mind, however, that this will also make it more likely for the user to dismiss results after this first set of profiled items as a lot less relevant than perhaps they are.

If you have a relevance ranking to go off and there’s a clear result or two that are far more relevant than the others, then why not profile just those. If several (more than three) of the top results are close together in relevance, resist the urge to profile any results. This kind of dynamic presentation (like presenting different content when you find no results – see above) can be used to the users benefit in various use cases.

4.      Results lack context

Okay, so I’ve gotten a whole list of useful results (and this time the information being displayed to me is readable and well laid out) – at least I think they might be useful results. In trying to determine whether I should click through, I need to know some information. There’s nothing worse than a search engine that shows just a list of titles with absolutely no context as to what page clicking that result might lead to. I’m hardly going to try everything your search engine suggests on the off chance.

So make sure to give your users context.

As a minimum, show them a title, and a summary of the target result (ideally showing your highlighted keywords in context). On top of that, try and show as much relevant meta information as possible. This include things like the date the page was published, the section of the site it’s in (products, about us, etc.) and whatever else might help me, as a user, decide if that’s the page I was looking for. An author, file type, the relevance ranking from your search engine, and even the URL of the page (let’s hope you have a nice RESTful site) are all possible additional items to show. While this is related to the previous point about use of real estate and good layout design, having the best layout in the world won’t help if you don’t have meta information to present. This of course brings us to the importance of meta data in search engines, but that’s another post entirely.

One item that’s often overlooked is the relevance of the match. If your engine delivers a decent relevance match indicator (perhaps a percentage), it might be worth showing this to the user. Let them see your engine is not sure about the quality of the result it’s found – or that it’s confident this result might be what you’re looking for. This let’s a user quickly judge whether the result is more or less likely to be worthy of their attention.

You don’t always have to show the exact relevance number, but some indicator to show the confidence in the result (one to three dots on a bar, a lighter to darker coloured circle, or some other simple visualization will suffice and can help). This kind of presentation is especially useful on a search interface inside an enterprise where you don’t want to waste knowledge workers time on the off chance that you have a match to their query. On the customer facing web, maybe it’s beneficial for a customer to look at a product that isn’t exactly what they’re looking for, especially if you’re not selling anything else that matches their exact need – in this case you might decide not to display a relevance result.

Make sure the results and their meta information are clearly laid out, and that the key items are larger and more visible than the supporting details. Detailed best practices for search result layouts come down to your user types, your content, the types of results being found, and many other issues. Best practices for different types of layout would be a whole other topic, but you can always look to Google to see how the basics can be laid out neatly.

5.      No guided search refinement possible

One thousand, four hundred and sixty results found. Happy clicking!

If your site just dumps back a series of results, that’s not usually that helpful. Unless your search engine is prescient, it’s not going have gotten the right result near the top of those results every time, for every user. It just isn’t. It doesn’t matter whether you spent millions of dollars on your search engine and the infrastructure, it runs on, or whether you’ve downloaded a basic keyword search tool.

Acknowledging this fact is important – if you think your search engine gets the results in the top few every time for everyone, then there’s nothing you need to do. Once you realize that’s not true however you need to think about how your users are going to refine and filter through the search results to find the one that most likely matches? Will it be through guided searching or refinement options.

This process involves offering the user options to further filter the results they’ve been presented with. As an example, perhaps the product class or document type is offered, so that the users can further slice their search results down to a specific category. In the best case, these filters are contextual and only offer refinements that are applicable to the search result set. Another example is that perhaps the search engine can extract keywords in the result set that are common or relevant. By providing the option to the user to add these additional words, the list of search results is paired down.

Almost every search engine has the capability of offering some kind of refinement capability. Even on the weakest side, you can probably direct the engine to match a keyword in the URL of the document and if your site is RESTful, the category or class of product will be in the URL so through this method you can always offer category filtering. Perhaps it offers the ability to match file types via the file extensions. Date filtering is also frequently possible. A lot of search engines do some kind of spelling corrections – make sure these kind of options (that don’t filter, but take you off in another possible direction) are also offered in the interface.

By presenting filtering and guided search options next to a set of search results over a certain number (it’s not usually worth refining less than 10 results, you can just read them all), the user is given a next step in the search or, in this case, the information discovery process. Your user interface can almost always offer such options, but frequently people don’t include them.

 

I hope this post makes you think about some of the issues that might exist on your web site or in your search tools, even if you’re not making these specific mistakes. What other problems have you seen in user interfaces for search? Do these kind of problems make the experience of search bad for you? Have you got other user feedback that suggests you have an interface problem? Let us know, we’d be pleased to hear your thoughts, thanks. 

Please note, I’ve tried to blue out any screenshots I’ve taken of sites that exhibit some of these mistakes so that I’m not picking on any particular companies or organizations. I’m sure if you wanted to you could figure out who these companies are, but that’s not the point, the mistakes are common, really. I’m sure if these search interfaces happen to be on your company’s web site you’ll recognize them. I hope you find this advice useful. Feel free to give us a call, of course, if you’d like any more input or help with your search interface.

4 Responses to “Top 5 Mistakes in Search Tool Design”

  1. Great post John. I completely agree with you on how the user interface needs to be intuitive well structured for the end user to quickly find what they are looking for. At nGenera we believe that its not only a robust and powerful Search Engine that is needed but also a robust and easy to use content management application.
    Intercepting the Search experience with a contact channel to reach out to the customer also makes for a much richer customer experience. Eg. Chat, Click To Call functionality can be embedded in a Zero search results scenario.

  2. John Tobin says:

    Nikhil – thanks for your input! It’s true that if the content you’re searching against isn’t great, search won’t help much. So making sure there’s a method to allow the business to deliver timely and relevant content or product information is always important. Adding other avenues for customers to get help from your company if search isn’t delivering what they’re looking for is also great – click to call on Zero results as you suggested would be a great example of making sure that user’s aren’t hitting a dead end.

  3. Shaun Ryan says:

    Nice post John – although I disagree with you on a couple of small points.

    In my experience relevance indicators are interesting to people who design the relevance algorithms & maybe to the people who install the search software – they are almost never useful to the end user. Especially if they’re not technical.

    Similarly, showing URLs in the search results is seldom useful. The end user doesn’t care what your URLs are and because the results are normally all from the same domain the first 25 characters or so are the same for every URL.

    The real estate on the search results pages is valuable – you should only show the meta information that is useful for the majority of the people searching.

  4. John Tobin says:

    @Shaun – great points, thanks. I was drawing on some of my experience here on intranet enterprise search interfaces where I think these things (relevance and URL) are often more helpful. I’ll address them separately:

    Relevance: Showing pure relevance calcualations may make little sense. However, Autonomy IDOL as an example gives a great percentage match rating that’s often useful. It’s most useful when it’s low – the engine is pretty frank about being clear that something might not match well, and communicating that to a user (basically saying ‘these search hits might not really help’), is powerful when you’ve given them other avenues (guided search, etc). I stress that this is only useful when the match can be communicated in a simple manner (percentage is as complex as you’d want to get, a simple bar graphic might be easier for example), and when the engine is good about telling you when it’s not sure about a match. If you’re a product site or have a basic search engine, I agree, DON’T confuse the user with this information. So good catch.

    URL: Again, you make a great point. The URL if it’s the same for the first part is not useful. I would leave it out in this case. I’ve seen enterprise platforms indexing content from multiple sources however and the URL helps distinguish sometimes. If you’ve got nice RESTful URL’s too, it can help the user determine – ‘oh, I’ve seen that result’. But you’ve certainly got to be careful (there’s plenty of ugly URL’s that are helpful to no one!).

    As with any information presented, it’s got to be useful for determining if the result is good for the user. Real estate IS tight (I originally titled one common mistake as ‘bad use of real estate’) and it needs to be used wisely for sure – in that we completely agree. Good clarifications to add to this post, thank you!

Leave a Reply