Friday, April 8, 2011

Context and Searching

In the course of thrashing over search results for our eCommerce site, a stakeholder suggested that we look into "context-sensitive" search. Now what did they mean by that? What is context?

"Context" is nothing more than the words and vision and words in the user's mind when they are searching. It's the motivation for and the desired outcome of a search. How do you deduce that when you're handling on-line queries? Do you make a guess based on previous searches by that user? Are those searches during the current session or do you consider past sessions? Or do you extrapolate based on the most typical searches by all users of your site?

The stakeholder quickly saw that we were already in deep philosophical waters so they offered a shortcut. "Let's just look at where the user is on the site (i.e, a catalog) and limit the search results to that section."

"For example, if the user is on menswear, whatever they search for, only show menswear results or at least, only show clothing with the menswear ranked first." Some people in the discusssion thought this sounded reasonable.

But let's think again about shoppers and how they behave. My first jobs during school and after graduation were in big department stores -- 10 years, and a lot of it spent on the selling floor working with customers. I observed several basic patterns to shopping that we can extrapolate to eCommerce searching. The first two we don't need to worry about; that's the expert who already knows their way around and has devised their own shortcuts; and the explorer who just wants to wander and browse for the sheer fun of finding something they didn't expect.

The rest of the shoppers want help and they have three ways of letting you know - the first kind comes in the door and immediately asks, "where can I find X?" That's it - "X" is the context. Nine times out of ten, "X" is nowhere near the door or aisle where the customer is now. All you can do, if you don't recognize what they mean by "X", is engage them in a little dialog and then direct them to some spot far away.

The second kind of shopper goes to the approximate area they are interested in and then asks for help to filter something down by brand or size or style or price point. Think of a lady standing next to the cold medications who asks where to find the aspirin.

The third type is a combination of these two. They've gotten what they want in aisle one and now they want to know how to find something that's in another department. They may be working a list that's in random order or some logical order known only to them, either way, their order may not match up to your site layout -- what they want may still be the equivalent of five aisles away, across the store, on another floor, etc. The immediate surroundings are not going to help in this situation either.

So 3 out of 5 search scenarios depend on getting help and 2 those cannot infer a solution from the immediate context (surroundings). What's a web site designer to do? All you can do is compensate for the salespeople who who were left behind in the brick-and-mortar world.

"Did you mean this?" (show list of similar and related terms)
"People who bought that also bought this."
"Would you like to look in this department first?" (show list of facets)
"Here are things you've bought before. Are you looking for something similar today?"

Unfortunately, those strategies mean you have to spend time collecting search terms and analyzing post-search traffic patterns. You have to collect histories of your customers and mine the profitable ones for patterns. And ideally, you can hire an information architect to optimize your navigation and map search terms to likely synonyms. All of which goes right back to those philosophical questions and the fact that fine-tuning search takes an investment in time and expertise. There are no shortcuts.