Improving on-site search functionality can benefit a site by contributing to a better user experience, and by reducing the barriers for users to reach their destination pages.

There are a variety of tools that you can use if your site does not already have a search functionality; I won't second guess which search application will best suit your site.

There have already been posts on various blogs about best practices for implementing site search (such as: label the button 'Search' and not 'Go', search case-insensitively by default, always recommend refinements to searches which generate no results, etc. Stoney deGeyter covered a lot of the important usability aspects last year.) In this post, I'd like to suggest some different techniques you can implement to get more benefit from your site search.

It is worth considering that, as with web-search, the type of queries made through on-site search can be grouped into categories of 'intent':

• Navigational: The immediate intent is to reach a particular page that the user has in mind, either since they visited it in the past or because they assume that such a page exists.
• Informational: The intent is to acquire some information assumed to be present on the site. No further interaction is predicted, except reading the contents.
• Transactional: The intent is to perform some web-mediated activity. The interaction constitutes the transaction defining these queries, and they are the most difficult to evaluate.

(Adapted from the concise descriptions supplied by InfoVis.)

With these in mind, we'll start by looking at collecting data to help you in this project.

Review Search Analytics

Not a tip, but the place to start is by collecting some data about the way users search on your site. Your analytics package should include a feature to monitor the use and effect of your on-site search. Google Analytics hides this under Content -> Site Search. Follow the site search instructions to expose your search query parameters to Google, and you'll be able to view a dashboard - such as that shown below from Mixcloud - showing metrics such as the percentage of visitors making refinements to their initial search, the average time spent on the site after searching and the percentage of searchers who left the site after seeing the search results.

If that wasn't enough, you can see the volume of each search made, those stats broken down by keyword:

Use Search Behaviour to Guide Site Structure

A simple review of this information can often give actionable items. In the example above, a lot of searches are for specific genres of music. This suggests that the users may prefer to find content based on a style they like, and the site architecture or navigation could be adapted to suit this behaviour. For example: a simple change could be to add a 'Genres' menu / tagcloud / etc - and populate it with the most searched-for terms.

User experience could be further improved by helping users get straight to the pages which receive the most navigational search queries - in this example by giving a front-page feature link to the mixes by Erol Alkan

Use Search Behaviour to Guide Site Content

A massive opportunity for many larger sites is to look at the search terms that receive high volume, but result in a high percentage of people leaving the site. In these cases, your users are telling you precisely the type of content or products (for e-commerce sites) they'd like you to provide! You can, and should, action this right away.

Using Constrained Search

Since search can be considered as a navigational tool that helps users to find the page they need in a more effective way than browsing through long category lists, sites which have a fairly strict site architecture can reflect this in their on-site search. Instead of having a 'free search' text box, they can have a number of fields which 'constrain' users to search in a way that matches the structure of the site.

For example, TrustedPlaces have a 'search feature' which asks users to enter a place type and a postcode.

This type of search form ensures that users are entering enough search information to ensure a quality result on the first search. If the results are disappointing (by being too broad, for example) then they will have to refine their search, or may simply leave the site.

Hijacking Search Queries

In many searches with navigational intent, users will benefit from being taken directly to a content page, rather than a results page. For example, a search on SEOmoz.org for 'ranking factors' could be improved by taking a user directly to the Ranking Factors page, rather than the search results page for that query.

The main SEO benefit of taking users to a content or browse page, instead of a search page, is that it encourages users to link to your well crafted page for 'widgets' rather than just the 'widgets' search results page - which is less likely to rank in Google and less likely to convert.

It wouldn't take long to do a review every week / month of the top hundred searched-for terms, identify navigational searches, and map these to the intended target page.

Have a unique URL for each search result

If your search results URL isn't unique to the search query submitted (e.g.: because you have used a POST form directly to the results page) means you could be missing out on the opportunity for lots of search traffic. Google typically avoids returning search results pages in it's own reults, but in many cases, the 'search results' are atypical and could be a relevant page to return.

For example, I find My IP Neighbors a very useful site. If their search page redirected to a URL that looked like www.myipneighbors.com/check/www.seomoz.org then they could well compete in the long tail of web searches for domain names.

PPC Landing Pages

One of my favourite pieces of social-research show that users searching for singular terms (e.g.: toaster) are further along the buying process, and should be sent to a product page, where as plural searches (e.g.: toasters) indicate that the user is looking for comparisons and responds best to being offered a range of options.

For people managing paid search campaigns, this means that site-search results pages are a quick way to generate a comparison landing page - and these pages typically have low bounce rate as users tend to visit at least one or two returned results.

The 'toasters' search demonstrates a lot of PPC campaigns using this quick and valuable technique, including sites such as MoneySupermarket and Lakeland Plastics. By contrast, Asda are using this technique to send 'washing machines' traffic to a page that reads "We're sorry but there are no results for your search" - please don't waste your PPC budget like this!

Using web search keywords

I think this is a brilliant idea for anyone who can apply it to their site. If the visitor has come from a web search engine, then you can pre-fill the search box with their search term. A very basic example of this on Youtube is shown below.

One of my favourite implementations of this was on Flickr. If you went from web search to an image page, the site-search box would be pre-filled, and a pop-up message over it indicated how many more images could be found for that search term.

For example, it would say "Search Flickr for 809 other images matching 'mexican wrestler mask'". This aims to keep people on the site for longer (and from not going back to web- or image-search) but for some reason, I've not seen this feature on Flickr for a while.

There's only so much that can be said in 1,300 words - if you have any particular questions about on-site search, feel free to drop them in the comments, and please do share any particularly creative uses and examples of site search that you've seen online.