Faceted Search Design – Ordering Facets

I’ve been lurking on the IAI’s mailing list for some time now, but recently someone posted a question that I just couldn’t resist answering:

We are getting ready to roll out a new faceted search option and I’ve been asked to make recommendations regarding the order of the facets and their characteristics.  I am having a hard time finding specific information about standards or best practices.  I repeatedly come across Stephanie Lemieux’s recent article, Designing for Faceted Search, stating that both facets and values should be based on importance.  While this is great, can anyone point me toward supporting information or is this something that is just understood?  Are there general guidelines for when to list characteristics alphabetically versus when to list them in descending order?

Ok, good call on the poster’s part – I had been vague in my article (Designing for Faceted Search, originally published in KM World), mainly because I had a broad audience and a word count limit. But I supposed that I should clarify a bit…  Here’s my response:

Re: facet order… There are no uber-best practices in terms of what order facets should go in other than “importance” simply because everyone’s context is different. In some contexts, there are some good rules of thumb… Like on an e-commerce site, product type should always be the first facet. In a multi-content/federated search context, content type, doc type or asset type is usually first (e.g. Libraries). Intranets often put departments of origin first. Beyond this, you’re stuck having to do some user-research to figure out what order facets should go (as others have mentioned). Given what they are looking for, what perspectives do users gravitate towards first for refinement? Is it subject? Geo region? You can get a sense of this simply by doing a few interviews. For a recent e-commerce project I did, we took each product type individually and had custom orders – for bedding, the size of the bed was most important (e.g. Queen, king, etc.), then the thread count, then the price, etc. I know the e-commerce examples are a bit tired, but you can extrapolate from this for other contexts.

Re: value order… This is a little more straightforward. To me, the whole strength of faceted search is in automatically presenting what is assumed to be the most meaningful choices first, and usually this is in decreasing volume of corresponding items in the results (a version of “popularity”, and typically the default for many faceted search engines). Typically, you are only showing the first 5-7 options of any facet at once, so it becomes more important to show the ones with the most bang at the top. If you do alpha, you risk showing options that only have a few documents associated to them and wasting that space or forcing another click for a larger number of people. So, I very rarely do alpha (I’ve done it once for Geographic regions, and once for brands, and only once the full list was expanded). If you are not doing true faceted search (with facet values automatically derived by the result set and numbers of documents for each showing beside the label) and doing something more static or navigational, then you can use more custom ordering (e.g. marketing value, putting better selling items first). Or if you are wrestling with a faceted search engine that doesn’t do decreasing volume sorts, then you’ll have to get creative and alpha may be a better default.

If any of you want to contribute more faceted search best practices, please do so in comments and I’ll compile them into a new list.

For more info, there’s a journal article by Marti Hearst (Design Recommendations for Hierarchical Faceted Search), or you can listen to Richard Tomlinson (Endeca) and I present on this webinar.

One Response

  1. I think you have hit all of the important points. The only other things I can think of would be:

    1) If you do show a subset of values within a facet (5-7 as Stephanie points out), it should be possible to expand the remaining values to see more of the values. If, say, the first 5-7 terms cover about 80% of the content tagged with that facet, it can be frustrating for a user to not be able to use remaining terms of that facet to navigate the result set.

    2) I think using the volume to order the facet terms is definitely the way to go. A minor nuance would be that if the terms have a mostly flat distribution of usage (and not a Zipf curve or similar distribution), an alphabetic ordering might still be useful (with the caveat that you should still be able to navigate to the remaining terms in that facet!)

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: