Summary: Consistency is a pillar of information architecture, but mixing classification schemes can be useful under the right conditions. As users, we think about things in flexible ways, and so our designs must flex too.
This is McCormick’s red food coloring:
Now, if I asked you to find this at the supermarket, where would you look? If you’re like me, you’d head straight for the baking aisle. And that’s precisely where I found myself just before Christmas, standing with a puzzled expression in front of a display of icing, sprinkles, and cake mixes … but no red food coloring.
This is a tale of classification and consistency. I needed the food dye for these red velvet peppermint swirl brownies, which my fiancée and I had planned to make for a holiday dinner with friends. It was the final item on my shopping list, and not seeing it anywhere, I walked back up and down the aisle several times, eyes scanning the shelves, and thinking “It has to be here, because where else would it be?”
Raise your hand if you’ve ever had that experience with a website or application, looking for something that just has to be there but isn’t. In this case, my problem was that it never occurred to me to search the aisle’s content by anything but its type — spices, stocks, mixes, and so on, each in their own place. But it took a friendly store manager to illuminate that the food coloring was organized by brand. It was with all the other McCormick products, across from the garlic powder and under the rows of peppercorns.
To me, this was absurd. Making exceptions or changes in the middle of a schema — in other words, organizing things inconsistently — seemed like a surefire way to confuse people. And on its face, it is. But a little research has also convinced me that when it comes to information architecture, mixing classification schemes can actually be both appropriate and effective. Here’s how:
When classification is supported by testing, and consistency follows usability
After all, what matters is whether a schema reflects the user’s mental model, not whether it adheres to some perceived ideal of how things ought to be organized. Sticking with the supermarket example, many stores have ethnic food sections (Asian, Mexican, Caribbean, and so on) that don’t fit perfectly into what appears to be the overriding schema. Yet those sections are useful because there are certain products that customers expect to find grouped by country of origin. I can only assume that when supermarkets decide to create a special section, they’ve done their homework. The same should apply to anybody designing websites and apps.
When classification supports multiple paths for users with different mental models
Offline, this is often inefficient — if I think of Sriracha sauce generally as a condiment and another customer thinks of it specifically as an Asian product, it’s wasteful for the supermarket to stock it in both aisles. And one of us is going to be irked when we can’t find what we’re looking for. Online, we enjoy the benefits of parallel hierarchies. If I were looking for my red food coloring on Amazon, I could’ve found it multiple ways:
Kitchen & Dining > Bakeware > Decorating Tools > Food Coloring
Grocery & Gourmet Food > Pantry Staples > Cooking & Baking Supplies > Food Coloring
I even could’ve found it in Amazon’s section dedicated to McCormick products, if that’s how I thought to search for it.
When what appears to be a mixed classification is actually undergirded by a hidden hierarchy
All the credit for pointing this out to me belongs to a user on UX Stack Exchange, the wise and anonymous unor. Users don’t necessarily see an architecture’s foundation. To borrow an example from unor, a supermarket might ostensibly be classified:
Fruits and vegetables
But the “real hierarchy” might look like:
So even when a classification appears to be mixed, it can have an underlying, unifying organization that makes it work consistently.
When mixed classification fails
Now, let’s loop back to my search for red food coloring. Mixed classifications fail when:
they don’t conform to mental models supported by user testing, and
they violate hierarchies that are otherwise clear, well-defined, well-labeled, and consistent.
I don’t know for sure how anybody but myself conceptualizes food dye. What I do know is that I was in an aisle where 99.9 percent of everything in it followed this hierarchy:
1. Section: Cooking / Baking Items
a. Item Type: e.g. Sugars, Chocolates, Spices, Frostings, Decorations
i. Brand: e.g. McCormick, Giant Eagle, etc.
In short, McCormick’s food coloring was classified at the wrong level in an otherwise clear schema, making it difficult to find. That’s poor design, or a mistake at the very least.
As another wise UX Stack Exchange user points out, there’s nearly never a “perfect” classification scheme. As users, we think about things in flexible ways, and so our designs must flex too. And none of this is to even get into non-hierarchical faceted schemas.
Under the right conditions, mixed classifications can be even more effective than uniform ones. I’ll certainly have my eyes open now for this architecture pattern, both online and off.