- September
- 28th
- 2009
Search-dominant interface design
As our interactions with the web are increasingly dominated by search, a pertinent discussion amongst the IxDA community leads me to explore the role of on-site search dominance and draw some conclusions regarding its efficacy in various contexts.
The IxDA discussion list can be a peculiar place. Amongst the job broadcasts, tissue-paper critiques of other people’s work, calls-for-papers, book clubs and speculative ramblings on the direction of our industry, very occasionally a great little debate pops up. Props to David Hatch for starting a discussion on search-dominance, hypothesising “People search first because that’s how they’re used to finding info”. The debate subsequently built up a head of steam when industry heavyweights Jared Spool and Peter Morville got involved, recalling memories of their good natured and informed search debate in 2001 (”In Defense of Search” – Peter Morville).
I’ve recently proposed more search dominance for a major client of ours and it was important for me to demonstrate that search should not be at the expense of category taxonomy or traditional navigation paradigms. Rather search should support and potentially lead the interrogation of the site by reflecting and responding to intents. Currently the data tells us that people use on-site search as a fall-back when traditional navigation fails; contentiously, Spool suggests this takes place off the homepage at deeper levels after people have exhausted the primary & secondary signposting, my understanding is that this takes place because people have arrived at a lower level first (vide infra).
As-ever with our line of work, there is no simple answer and the caveat ‘it depends’ must be applied. Looking at the nature of the content will provide direction. Search-dominance is strong on Amazon and eBay: the breadth of product offering in these environments suggesting that a design dominated by categorisation wouldn’t work. Or would it? CraigsList recently underwent a theoretical redesign thanks to Wired and in one instance where search was presented as the dominant element, user feedback was interestingly negative. The breadth of taxonomic categorisation on craigslist and the familiarity of this for their users does lend it some support. I regularly cite Globrix, a British property search site that originally depoloyed little more than a search box on their homepage but has since modified their design to cope better with the a broader range visitor motivations (returning, non-buyers, agents). On Globrix the principle user-type (a property hunter) knows that the dominant meta data is location, this is their primary intent – find a house in a given location, only then will they filter on additional faceted needs such as price, dimensions, bedrooms and so on.
For the examples above the functionality and content of the site is well-known and thus either categorisation or search – if comprehensively considered – will be largely successful. Where the content is less well-known and Jared’s so-called ‘content identifiers’ don’t exist, search dominance might not be the best solution. Take the cited Sequoia Capital, as Hyperlabs‘ Stefano Bussolon indicated in the discuss, Sequoia has limited scent; one has to play with the search to uncover the information the site contains and this doesn’t particularly sit well with the nature of the content Sequoia holds, it doesn’t have the familiarity of Craigslist, Amazon & eBay. It doesn’t make interrogating their material impossible, it just makes it a few degrees harder and that just doesn’t seem right. Of course, there are other marketing reasons to do this to your site: a dramatic interface shift provides something to talk about, and gives you a great opportunity to fiddle and fettle with your content to start hiding or promoting various pages in-line with your business objectives.
Context of use should play a part in your thinking too. Bokardo’s Joshua Porter rightly directs us to consider the screen size of mobile users and the value of focusing on search where bordering a page in with navigation elements crowds the interface or directing people to scan and click within a tag cloud provides gestural/interaction problems.
Sitting above all of these considerations is the all-seeing eye of Google. We must accept that people are increasingly arriving deeper within (y)our site(s), Google doesn’t just send people to the homepage. People must engage with your architecture throughout the site and this may mean that – without the correct wayfinding at the point of arrival – they will resort to search from deep within the hierarchy. As William Brall analogised
“[your homepage is not] your front door… all your content pages are. More accurately, Google is your front door … or really your hallway.”
Finally, search can afford us valuable insight to what users are looking for. The interrogation of search logs and the careful consideration of user’s interactions with the search interface reveals – often in real-time – information about what content people are looking for and, importantly, how they categorise it. You may wish to expose this data back to your users “most searched for…” or use it to re-categorise and prioritise your content behind the scenes.
On-site search has an increasingly bright future, as the capability of the engines improve so will the results. As information architects we need to be mindful that taxonomies are, through their origins in the ambiguities of our language, inherently subjective. Exclusive reliance on categorisation is not a solution and that supporting them thorugh well-considered semantic and faceted search implementations will yield greater user-satisfaction and important insights to user-intents that can be gathered in real time to improve the findability of your content.
Useful Reference: Konigi’s pattern library of search interfaces





