nicolas leroy

Faceted navigation: how data quality matters – the case of product navigation

July 08, 2009

Faceted navigation (also called “guided navigation” or “parametric navigation”) is a design pattern frequently used by shopping engines; it’s indeed a convenient way to refine a search by applying filters. There is a good introduction of faceted navigation on; and a really interesting article on User Interface Engineering to learn how to design good faceted search. What those articles don’t clearly mention though is how the quality of data is important to build convenient and useful guided navigation for users.

To illustrate this statement, let’s take the case of “product navigation”: faceted navigation is used to find the best product that matches user’s needs by helping this user filter on product features; once he has found a product that fits his needs, he clicks on this product (often clicking on a “compare prices” button) to get all shops selling this product (then landing on a product page).

In terms of data management, I think the main difficulty to build an efficient product navigation is how to define products in such a way the price comparison is meaningful.

The camera example

User1 (beginner): “I want to buy my first entry-level SLR to learn photography. I want to buy a bundle with camera + one or two lenses with the best price. I’m pretty brand agnostic”.

User2 (expert): “I’m looking to buy a new SLR. I already own a Canon SLR, with a few lenses. I’m interesting to upgrade my body camera, but I could be interesting to buy a bundle in order to have better lenses. The price of the body only vs bundles will be key for my decision.”

User3 (expert/professional): “I want to replace my old camera with a new one; I’m happy with my current lenses and I’m not interesting to change them or have a new one”.

Let’s take US as an example for digital camera search. A “canon EOS 50d” search returns several results: the “body only” option as first choice, then several variations of the product with bundled lenses. The price ranges of those four “products” clearly overlap.

For User1 and User2, this is a tricky situation: it means comparing each product, see which merchant provides the best price… However, for User3, it’s a pretty straightforward experience. US SRP for canon eos 50d

The fridge example

User1: “I’m looking to buy a new fridge. The price and features matters. Classic color for me”

User2:  “The color drives my choice, as I want my new fridge to match with the colors of my kitchen

Let’s look at PriceRunner UK and perform a search on “smeg fab28”. Smeg is fridge manufactuer that provide its models in a wide range of colors (the FAB28 is available in 14 different colors including an union jack style :) ); adding the fact SMEG uses different model names if the fridge door opens on the left side or right-side, the number of combinations is stuning.

If I’m User2, I’m really happy with this product definition. If I’m User1 and the FAB28 (among other products) matches the technical features I’m interesting in, I’m lost as I have too many similar references I’m not interesting at the moment to take any decision…


YouTellMe, an implementation that works

YouTellMe is a quite new price comparison site, that I talked about a few months ago, currently launched in Netherlands only. When it launched, I was impressed by how detailed their product descriptions were and how such gap in data quality – compared to other CSEs – really improved the user experience. Recently, the site added the support of “product variations” that is a smart solution to the problems explained above.

A search on “Canon EOS 50D” brings one single result, with a quite large price range. A link “there is 8 variations for this product” (may not be the literal translation, my Dutch is inexistent) allows the user to show all variations of this product. In the case of this digital camera, variations are different bundles with lenses.

Clicking on the price range leads to a product page where it is easy to include or exclude variations of the product. I’m not interesting in any lens, then I choose the “body only” variation. I want to see if there are some interesting bundles, I include all the variations.

It works the same for fridges: fridges with different colors are considered as variations (see the Smeg FAB28 example).

The user interface of YouTellMe may be a bit too complex for the average user, but it shows we can still innovate in product comparison.

Conclusions / Thoughts

2 commentaires

on Jul 09, 2009 / 3pm
Agreed. It is frustrating to narrow a big set of products and realize you're missing key filters. Then what can you do? You can't take those results elsewhere for further filtering. In some cases you can use text-search, but it is only suited for some attributes, and the data still needs to be consistent. Usually you're stuck with the old-fashioned way, manual filtering, defeating the point of comparison engines. I'd like to see more sites focus on deeper data. The fact that a comparison engine features cameras AND refrigerators means little to me while I'm searching for a camera. And it's a negative if the time spent on refrigerators takes away from the depth of camera data.

on Jul 09, 2009 / 4pm
@Sean - The difficulty to be specialist vs generalist... Sure covering many categories drives you away from managing deep data, as the cost of maintaining data becomes huge...