In the web business, when you talk about search most people think of SEO and how to get more traffic from Google. Or Bing if they work for Microsoft. Or Ask Jeeves if they think they’re being funny.
Yes, I think I’m being funny. There’s a reason I’m a project manager and not a comedy writer.
However, most clients and web-builders don’t focus too much on on-site search. In most standard informational-style websites the focus is on creating a navigation structure that allows a user to intuitively find what they are looking for. On-site search is just the little box that goes in the upper right and connects to your platform’s built-in search functionality or an external service like Google’s Site Search.
This is is why I was interested in a recent finding from the Nielsen/Norman Group that improving employee search on your intranet can save your organization half a million dollars.
To be clear, they are talking specifically about employee intranets at companies with 10,000+ employees. So, you know, your mileage may vary.
However, there are some key takeaways that I think can be extrapolated to site search in general. These principals make up what I think of as “smart search” – a much better way for users to find information on your web site.
- A unified search field is better than multiple, specialized searches. If a user needs to use one form for a general text search and a separate form for something specialized like a contact lookup or product search you’re decreasing efficiency and increasing the likelihood of human error.
- A search box should be simple. The user should not need to know an advanced syntax or use additional form fields in order to get the results they’re looking for. Smart search tools should be smart enough to match different content types automatically.
- Results should display immediately. On most sites, search still functions with the pre-AJAX style form postback methodology. Some sites have an autosuggest feature that provides common search keywords as you type, but a good search shows top matching results as you type, with a “more” option at the bottom. This provides much more information to the user as to whether they are on the right track in the search task.
- Results should contain pertinent information. The example used by Nielsen/Norman is that employee search results should include a job title and photo for identification purposes. The results should also include some contact information, as this is what the searcher is likely looking for. Allowing a user to see the contact information directly in the autolookup results saves the step of having to click on a name and get to a separate “employee” page to find a phone number.
- A unified results dropdown should show different content types. A site might offer products, have a people lookup, and general pages about the business. Each of those types of content should display different information in the results (price and availability for products, contact information for people, and page title for pages) but still exist in one results dropdown, or on one results page.
Unsurprisingly, the sites that do search well are the sites with too much information to expect a user to use traditional navigation to find what they’re looking for. These sites also tend to have various types of specialized content, rather than just a collection of pages.
The most prominent examples of smart search are on social media sites like Facebook and LinkedIn, or well-funded database sites like IMDB. Surprisingly, I could not find any eCommerce site that was employing a smart search strategy. Most of them offered an autosuggest for search terms, but I would expect an autolookup displaying top product matches (showing price, availability and rating), top video matches (showing year released, maybe director and top stars) and general page results as I type.
Of course, there are huge implications to implementing a smart search box on your website.
- Planning: A smart search solution really needs to be properly planned. User testing and analysis needs to be used to determine what users are likely to search for and which pieces of information they need. Also you need to plan how search results are prioritized. What makes one result “more relevant” than another?
- Design: Currently few people think of “designing” a search solution. A smart search needs to be designed so it doesn’t just become a giant haystack to hide all your needles in, making it less usable than the old-fashioned search its replacing.
- Implementation: Not only does the search and display functionality need to be programmed, but there may be a good amount of work to turn unstructured data (web pages) into structured data (custom content types).
- Performance: Can you build a solution that will not affect the initial load time of a page or have a slow response time that frustrates users as they type? Keep in mind on the web “slow” can often mean fractions of a second.
- Customization: At this point even the big players have a way to go on their search UX. You can guarantee that there’s not going to be an out-of-the-box plugin for most content management systems that does this work. This means you’re talking about custom programming and all of the costs and dependencies associated with that.
These implications add up to one big explanation for why smart search is not more prevalent already. Creating a smarter, better on-site search is not a question of simply upgrading a plugin. It’s a project in itself that can be as resource and time intensive as any major application feature.
If your business model is not based around user search, its easy to make this a lower priority feature that keeps getting tables “for the next phase”. However, the great thing about the web is that even though annoying trends come and go, things that are better for users tend to win out in the end. It may not happen overnight and it probably won’t be as disruptive as responsive design, but I think that the principals of smart search will be commonplace on web sites and applications a few years from now.