ia play

1. Ranking

This is the least visible thing, that you might not consider a feature, that mostly gets ignored and is absolutely the most important thing for you to dedicate time to getting right.

If the query isn’t particularly ambiguous then you need the top results to be right, without asking the searcher to do much else.

Ranking isn’t sexy and it takes care and attention. But isn’t magic, it’s just rules. Ask what the rules are. Don’t be fobbed off. If no-one knows, work it out yourself.

2. Manual Suggestions (query expansion/narrowing)

This basically means Best Bets.

I’m very, very attached to Best Bets. This is mostly because I’ve been a search product manager as well as an IA on search re-design projects. Once the project team has packed up, the product manager (or web manager/editor) can still improve results and resolve problems using Best Bets. And they will need to. Promise.

3. Automated Suggestions (query expansion/narrowing)

We can’t spell and we can’t type. And then we blame the poor old search engine when it doesn’t find what we were looking for.

Any decent search solution needs to have some solution to misspellings (where to put them is a problem for another day!). You can do some of this with Best Bets, but with a big and diverse enough set of users you’ll probably need something a bit more automatic like Google’s Did You Mean?

A related but broader concept is suggesting related searches. You might have spelt your query correctly but there’s a similar term that would get you better results. Ask.com used to do this.

It might seem perverse to prioritise the manual intervention over the automated one. I’d usually expect to have both but I have a few reasons for picking manual if it comes to a choice:

• the manual option is probably cheaper to add on if neither comes as standard
• automated suggestions often get better over time but might start a bit ropy
• automated suggestions may be ‘black-box’ you might not be able to do anything with them if they are wrong/misleading. And every system I’ve worked with and/or used makes mistakes sometimes.

It’s worth asking whether there is any control over the automated suggestions. Is there a dictionary? Is the right language (esp. UK v US English)? Can we edit it? How?

4. Filters and sort options (after you got search results)

These tend to get missed by users or interfere with their understanding of the page. Not all users will understand them, especially complex faceted filters. The positioning of filters/facets is very difficult to get right. Users home in on the top results, so above the first result is most likely to get noticed and also most likely to get noticed for being in an annoying position.

If you are doing product search then I’d probably still prioritise 1-3 but I’d strongly argue you need 4 as well.

5. Clever query language

Quote marks seem to be reasonably widely understood, so I might argue these should be higher up your expectation list.

But unless you’ll have access to your users and be able to train them all… I wouldn’t prioritise operators like wildcards, NOT/And/Or etc..

Find out what you get out of the box. Make that information available to interested users. But don’t invest lots of development effort and money here.

6. Filters and sort options (before you run the search)

a) Radio buttons and drop-downs.  These get missed, people don’t think about using them, they tend to just stick words in and hit go. Other users won’t use them because they don’t know they need to use them until they see the search results aren’t focused enough. So then they have to go backwards. So you might as well go with (4).

If you can sensibly default them then they can be more useful but establishing what the sensible default  is problematic.

b) Advanced search pages.
These are basically a collection of filters for the user to set before you run the search. Search specialists inevitably find advanced search useful but your average end-user doesn’t. The exception here is power users  but be sure the users actually are “power” users.  You are likely to find power users where there are time/cost pressures around searching e.g. staff answering customer calls or researchers using databases where they pay for searches. In these situations even reasonably techno-phobic users are motivated to get to grips with advanced searches including some of the more complex query building ones.

Another reason advanced search might be worthwhile is if your power users are also your most mouthy. If the segment of your audience that blogs/tweets is also the segment that might demand power features then you might consider the feature as marketing.

(Don’t be worried by people being intimated by the label “advanced”.  If they are intimated by the word then they’ll be intimated by the features. )