I got to thinking about this question actually at the supermarket recently. I was looking for an item that I thought just had to be in the baking section ... because that's how brick-and-mortar stores are organized at the top-level, by type or category ... when in actuality this one item was an exception in that it was shelved with all the products of the same brand.

Are there examples related to information architecture where mixing classification schemes like this is appropriate or effective? A search led me to this article by Donna Spencer, in which she runs down various schemes (e.g. alphabetical, by task, by subject, etc.) and concludes:

In reality you can use a combination; you could:

  • Mix up types at each level
  • Start with one type and use a different type at the next level
  • Use more than one approach for your whole content set

It's not unusual to see schemes that vary by level, but what about that first bullet point? I'm hard pressed to think of an IA that suddenly switches classification and then switches back between exceptions. That'd be a navigation experience without any consistency.

Mixing schemes within hierarchy levels just seems like a flat-out bad idea. No?

4 Answers 4


Mixed classification in IA can be both appropriate and effective.

To begin with, a proper IA should be determined by user-testing. Activities like tree-sorting will reveal that users sometimes expect elements outside the main classification scheme. Generally in UX, consistency should follow usability (a statement that is the source of debate I'd rather not open here).

One example for this is the Polish section in UK supermarkets - whether it is bread, drinks or deodorants, these all appear on the Polish section shelves, rather in their corresponding 'native' sections. Polish buyers know where to go, and British buyers don't get 'Polish product noise'.

In addition, there's nearly never a 'perfect' classification scheme. Elements nearly always have a few facets, and often the 'logical' solution can only be derived by user research. For instance, the BBC has a top-level section called 'TV'; one of the BBC channels is the children channel (CBBC); yet CBBC also has radio and interactive content. So there is also a top-level section called 'CBBC', most likely this was determined by user statistics.

So if users is anything to go by, IA may work much better if it does not adhere to a strict classification scheme.

  • 1
    In addition, a polyhierarchy isn't uncommon on complex websites and applications. You're likely to find the same Black Flag album under "rock" (a broader category) and "punk" (a narrow category).
    – erik_lev
    Dec 27, 2013 at 7:25
  • 1
    +1 to the answers here, especially the points about consistency following usability and multiple navigation paths for different users. In the example of the Polish grocery section, tho, isn't that really still the same classification scheme? It's topical or type. Granted, it's more faceted than hierarchical -- a sausage can be "meat" or considered "Polish" simultaneously but both labels answer "What is this thing?" Another scheme might answer "Who is this for?" "When was it made?" "What can you do with this?" Etc. In short, am I even thinking about "mixed classification" in the right terms?
    – in_flight
    Dec 27, 2013 at 15:34

Isn't that the whole advantage of an ecommerce site over, for example, a supermarket. You can create different paths to the same product without creating duplicate product locations, suiting the customers who have totally different perceptions of where to find the product. If I would want gluten free bread, in a supermarket, I would search for it at the bread section. Somebody else would search in the products stand where all gluten free products are. In you're average supermarket either one of us will be looking at the wrong place for supermarkets won't place the same product at two different locations.

Another personal example as to why I think it's not a bad idea is when I was looking for a new TV and other smart media products. I wanted a TV with which I could easily communicate (preferably wireless) with my laptop, tablet, phone, sound system, gameconsole and what not. The webshop I chose to do my inspirational shopping at had an entire section devoted to Sony, promoting the fact different Sony product could communicate with each other without a hassle. I could find tvs, soundbars, mediaplayers, gameconsoles, laptops, tablets and phones. I was dearly tempted into buying just Sony products regardless of any product specifications and reviews.


I’d say that this would be no hierarchy in the first place. Or: A different hierarchy with unknown/hidden levels.

Let’s look at some examples.


A library might have sections/shelfs labeled:

  • Law
  • Medicine
  • Philosophy
  • For sale
  • Magazines

But "For sale" probably contains books that were formerly found in "Law". And "Magazines" contains media that also would fit into "Law". So what seems to be the top hierarchy level contains mixed schemes. But maybe it only seems to be the top level. The "real" hierarchy could be:

  • (Borrowable books)
    • Law
    • Medicine
    • Philosophy
  • (Saleable books) For sale
  • (Borrowable magazines) Magazines


A supermarket might have (unlabeled) sections that users would probably label as:

  • Fruits & Vegetables
  • Meat
  • Bakery products
  • Breakfast products
  • Nonfood
  • All products by ACME Inc.

But the "real" hierarchy might be:

  • (Entrance area)
    • (Perishable) Fruits & Vegetables
  • (Central corridor)
    • (Medium shelf life) Bakery products, Breakfast products
  • (Red light area)
    • (Chilled distribution depot) Meat
  • (Spotlight area)
    • (Sponsor) All products by ACME Inc.
    • Nonfood


A website might use the following navigation:

  • Questions
  • Tags
  • Users
  • Badges
  • Unanswered

Well, "Questions" and "Unanswered" on the same level?!

But the "real" hierarchy could be:

  • (Question lists)
    • (Questions relevant to the user) Questions
    • (Questions with no upvoted answers) Unanswered
  • Tags
  • Users
  • (Help)
    • Badges

Or it could be:

  • (Most important links for our users)
    • Questions
    • Tags
    • Users
    • Badges
    • Unanswered

You can come up with countless hierarchies for everything.

If it seems that a hierarchy level contains things that don’t belong together, this level probably has a different label than assumed.

If something is clearly intended as hierarchy, and the labels are given explicitly, then yes, I’d consider it a very bad idea (an error even) to add childs that don’t fit the label. Example:

  • Shoes (by color)
    • Black shoes
    • Blue shoes
    • Red shoes
    • Reduced!

"Reduced!" clearly is not a color, it doesn’t fit the label.

That doesn’t necessarily mean that this link wouldn’t be useful to visitors at this position. But it shouldn’t be designed as being part of the color categories.

  • This was a particularly illuminating response, so thank you for taking the time! Hadn't thought of the whole notion of hidden hierarchies versus the hierarchy the user actually sees. If I wanted to research more about that, I don't suppose you've got resources you'd recommend?
    – in_flight
    Jan 4, 2014 at 13:48

Short answer.

Faceted Navigation, these where designed to solve this problem.

If you want to mix task with subject you can. Even a descent sized online shop is rendered difficult to navigate without using facets.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.