10 thoughts on “Faceted Finding with Super-Powered Breadcrumbs”

  1. A thorough, careful, and well researched approach to the challenges surrounding browse and search. IFB seems dramatically more intuitive than traditional breadcrumb solutions I’ve encountered. Thank you.

  2. Hi Draka —

    No money, decorous or otherwise. I just wanted to give away my best design to date as a kind of social experiment. That said, there are many critical insights that simply would not fit in the article, so if you are thinking of implementing something like this, let me know: Greg@DesignCaffeine.com

    Glad people are finding this article useful — thank you all for your comments!

  3. I think this is awesome. There are a few content heavy sites I work on that would seriously benefit from this.

  4. Hi, Greg,
    This is absolutely insightful and valuable. Thanks for sharing this, complete with user research on it! One thing I particularly like about your interface design here is that the user can see both the facet and the facet value in the breadcrumb trail, as well as being able to select a different facet value.

    As I read this, a couple of questions came up (maybe I’m slow!).
    * In your discussion of Change instead of Set-Remove-Set, the Change example allows user just one choice, and making a different change updates the results I assume, but the Set-Remove-Set allows user to select any number of choices. In the big Walmart example of IFB, you show a dropdown of checkboxes and a Go button. That made me think that I don’t understand your intent in this discussion. What’s the differentiation you’re aiming for here?
    * In your Figure 9, you showed multiple checkboxes as a user option; if you allowed a user to select both Nikon and Canon, would you display both facet values in the breadcrumb trail? Or…?

    Thanks! Christine

  5. Very interesting work! Especially since the concept of “change” seems so much more intuitive than “set-remove-set”. I;m intrigued and would like to start using some of these concepts, but have a couple of questions about how you would propose handling some of the stickier situations that can arise:
    1. Does this model work with multi-select dimensions? That is, does this model still work in a situation where the user is allowed to select more than one value per facet?
    2. How do you handle hierarchical facets like category? In the examples shown here, you only show a single level of category represented in the breadcrumbs. Do you only display the lowest level reached and provide access to select or change the higher levels behind the drop-down?
    3. When a user decides to go back and change a previously selected facet selection as in your example for figure 9, do you provide the full list of available facet values for that facet for the entire data set or do you constrain the available options? In the example shown, selecting lenses would be outside the current set of results since you have already selected to constrain by 12Mpx, which implies that you are covering the entire scope of available options. This can lead to unintended consequences like zero results unless you drop other conflicting facets, but how do you decide which ones to drop and in which order to avoid the dead end?

  6. Hi Christine and John — thank you both for your comments.

    John: “…this can lead to unintended consequences like zero results unless you drop other conflicting facets…” I addressed this exact issue in: “3. Automatically retain relevant query information” — yes, I recommend you drop conflicting facets. As to what algorithm to use to select the facets to drop, it will depend entirely on your system: some combination of visitors, conversion, available inventory and category could all be considered. My background is in software engineering, and I can definitely help you build the best algorithm for your needs.

    Christine: “if you allowed a user to select both Nikon and Canon, would you display both facet values in the breadcrumb trail?” How the OR is implemented, and whether you need it the OR function at all will depend solely on how the system is to be used. For example, it’s rarely useful to show two categories joined with OR, but atomic values like sizes and colors quite naturally fall into OR paradigm. I discuss some of the filtering challenges here: http://www.designcaffeine.com/2010/articles/321/ and these challenges certainly apply to IFB. Plus, there is a whole question of using a “Go” button vs. an Ajax call to support multiple-select… Again the answer depends on the business and customer goals you are trying to achieve.

    Based on high interest of implementing IFB, I am doing a series of 1-2 day workshops at client’s locations designed to address the specifics of individual systems and ensure successful implementation. Please feel free to email me at Greg [at] DesignCaffeine.com to see how we can continue the conversation.
    Thanks again for your comments!
    – Greg

  7. Greg,
    This is brilliant. The mental energy you save the user can be devoted to finding what they are looking for,
    not fighting the query engine.
    Thanks for sharing this with us.
    Lloyd Muccio

  8. Great article! I’ve been building websites based on Endeca.com’s software for some years now. The latest project is based on Lucene, though. At the beginning I had to convince usability agencies that Guided Navigation may be a good idea. I’m glad the concept has evolved so nicely, even Google is using it by now 😉

    I will speak on Faceted Search this Thursday at a Swiss University http://e-byz.ch/usability/guided-navigation-faceted-search-websites


  9. Thanks! This information is exactly what I was looking for. I work on a lot of information sites that contain thousands of item that don’t necessarily fall into one neat spot.

  10. Nice article.
    Do you know of any site that has these ideas implemented?
    It did not find it on any of the sites that you mentioned (no idea if you were in any way involved with them or not).

Comments are closed.