How internal search should be working on your site

Let’s split search first of all into two buckets “External search” and “Internal search.”

External search – Google wins, let’s move on. Internal search is something worth talking about. In fact it is often overlooked just how often such a feature is used. Jupiter Research conducted a study in 2006 that concluded that 34% of visitors use a search function as the first thing they do when landing on a site. It also concluded that 82% of retail site visitors will use a search function during their visit. Forrester Research backed this up in March 06 with a finding that 50% of users surveyed used search as their first port of call on a website they had not previously visited. Internal search is getting more of a workout than I would have imagined.

If internal search is working right it can become the ultimate navigation tool. As people have started turning their attention to the “mobile Web” I think this trend is set to continue. Voice recognition on mobile devices combined with site searches that work may actually give the web on the go a chance of success.

Surely with this level of importance a search box has a rightful place in the top level navigation. It makes you wonder how both Ford and Holden missed it. Toyota managed it though.

The opposite of no search is complicated search. I’m going to pick on the search interface provided by Snitz Forums as can be seen on a South Australian EPA site. Complicated search is often accompanied with options like AND OR and EXACT MATCH. It often allows for some kind of filtering of results before a search is made.

Filtration is good, it can help you narrow results and potentially get you where you want to be. But I believe it is better to allow the user to search first, try their luck, they may find what they want without having to answer another question. The best spot for filtration questions is over the results of a search.

Ok, we have moved from search input to results, one has to wonder about an institution that can go to the effort of integrating search input into a website; go to the effort of getting an indexer installed and scanning content. Finally, putting development energy into hooking the search input to search results. After all that effort, you would think a bit of testing of the results would be in order.

It is Mitsubishi Motors turn. Fuel efficiency would be an appropriate thing to search on I would think. Toyota does a good job on that term, but I can’t post the link because they stupidly use a POST request not a GET request for the query (somebody needs to read up on REST). Mitsubishi however returns nothing. I find it really hard to believe an Australian automotive site not mentioning fuel efficiency at all. Oops, turns out I am right. I am not surprised that Google can “out search” an internal search engine. Until Microsoft made a concerted effort with Live search Google was “out searching” even them, anyone doing software development and using MSDN online as a reference during that time can testify!

So Mitsubishi needs some help with their search engine’s capabilities. Let’s keep trying for some results. How about just “fuel”? That had to return something but wow what underwhelming display. No abstracts is an interesting choice. Just titles and full URLs, I’m not sure URLs that long are particularly helpful to the end user especially highlighted the way they are in red. Even better, the black title is the navigation element; the highlighted red URL does nothing. I find that really confusing.

So far this post has concentrated on document search otherwise known as unstructured search. It gets the unstructured name from the fact the “corpus” (content being searched over) is not a database that does have structure.

Unstructured search is generally left to off the shelf style software. To sample a few there is Indexing Service from Microsoft; Lucene from the open source crowd; there is even an interesting hosted service from picosearch.

Structured search is done on things like employment sites, used car sites etc. And require a custom software development approach.

Many sites in this camp have only one sort of content, even so, the concept of starting a search free form and then allowing specific filtration controls that are built around the type of structured data being searched over seems like the smart way to go.
If you ever find yourself in a situation where there is a multiplicity of structured content to search over, well, don’t find yourself in that situation – but if you do http://base.google.com is where it is at!

Another common occurrence is the need to integrate document searching with structured searching. Again, I would recommend a single simple search box followed up with splitting the results into categories if necessary as a type of filtration device.

Points of the post:

• Search is used more often than most people think
• Simple search input ideally as a persistent navigation element is a good idea
• Test your results!
  

Happy searching!

All "A Tag Short of Compliance" Blogs

In his real job, Luke Amery works on shopping cart software. He is the technical director of On Technology, Australia's leading e-commerce development company.