What's the best search for everyone?
One issue we often think about is “what’s the best default search type for everyone”? There are a few design constraints around the problem. The biggest one is that users have been educated (!) to expect a single text input and a single search button and from that input we want to produce as relevant results as possible. With those constraints it’s genuinely hard to come up with a default search that is both meaningful for skilled search professionals, but also doesn’t overwhelm the average user.
To be clear, I’m not talking about typical “advanced search options”, like exact and part word, wildcards, similar and vowel and numeric replacement. We support all of those and more. But they’re (rightly) provided as options for skilled users.
And it’s not a problem that can be solved by “popularity” or frequency based algorithms that general purpose search engines might use to locate documents. While we track trademark popularity, even an unpopular trademark can be sufficient to prevent use of a brand.
The question is: What search you do from a single text box and one click on a button?
In one camp is “exact”. It produces great, simple to understand results with razor sharp focus. When you’re looking for a known trademark, this is the one to use. It’s perfect for our domain name and social media username “availability” search results (which are also on the same page). Obviously you need to provide exact as an option, and we conform to the expected standard by putting double quotes around the relevant phrase. But trademarks are obviously more complex than an “exact match” search, and we can be smarter for novice users.
In another camp is “starts with” (normally with some ignoring of spaces, hyphens, etc). It’s wider than “exact” so grabs more results that are relevant. It’s also easy (computationally speaking) on the database. But “starts with” really isn’t much more relevant than “exact” in that both could equally miss relevant marks and it bombards the user with more, less relevant results too.
We’re in a third camp. At the moment we default to a proprietary search strategy, which combines a number of advanced search options, data from outside of the trademarks register and additional filters – let’s call it a “smart search”. It’s not as creative as our “like:” search, but it’s not as limited as exact or starts with. It’s a balance between grabbing the most relevant results for the novice users without overwhelming them and satisfying the skilled searcher who wouldn’t live with an “exact” or “starts with” style search. The issue then becomes educating people on what search you’re doing – and we’re purposefully keeping it close to our chest! So while the search might be really smart, a skilled searcher is likely to go back to the normal “advanced” search strategies so they know exactly what search has been done. That being the case, perhaps we should too?
We think innovation around search is one of the most exciting areas in information technology. Search has been great at organising massive amounts of otherwise unorganised information. But it’s still got a long way to go before it can work out the intent from a single line of text and one click!
We’re really interested in thoughts from the community.