December 30th, 2008
My article, Design Cop-out #2: Breadcrumbs, is one of the most controversial I’ve written in recent years. People either agree completely or think I’ve gone off the deep end.
When people disagree, it’s often because they think I’m suggesting that we stop putting breadcrumbs in our designs. I’m not suggesting this at all.
I’ve defined a design cop-out as something that happens when the designers focus on treating a symptom instead of addressing the root problem. A cop-out is a red flag that should be raised in the design process, to ask the question, “is there a better way to solve the problem?”
At doteduguru.com, blogger Michael Fienen wrote a thoughtful rebuttal to my article with many of the questions I often get when I start talking about my thoughts on Breadcrumbs. Responding to Michael’s points makes for a nice way to talk about these issues, so I thought I’d take some time to do that.
Surfacing the Content
In the original article, I said:
The idea behind how breadcrumbs should be used is simple: the user ignores them until they get to a page that isn’t quite what they wanted. They discover the trail of links and click on the one most likely to contain the correct path to what they were originally seeking.
To which Michael responded:
I think [this idea] is patently incorrect. A user doesn’t necessarily click on a bread crumb because they think it will take them somewhere better or put them on a correct path, nor is there any reason to believe they are used only by lost visitors in the first place. They click them so that they can surface up in a web site and potentially begin navigating anew.
Micahael’s not the first to suggest this. Many information architects I’ve talked to hold this, as we see it, common misconception: breadcrumbs are not only a loss-recovery mechanism—they also serve as a tool for “surfacing the content” of the site.
What’s interesting is when we’ve studied users, both in the lab and in the wild, we almost never saw them interested in “surfacing the content” or learning more about the site. Sure, they want to find the content they desire. If the target content is on more than one page, then they need to get to the subsequent pages. But beyond the user’s explicit target content, we never see them show any interest in the other available content on the site.
Since our early studies on the web, more than 12 years ago, we noticed that users are always on specific missions when they come to sites. With only one exception, users never visit a site “just to see what it has.” (The one exception? Web designers.) They always have a mission:
- Buy a new winter coat and accessories
- Find out what my portfolio is worth
- See if my favorite blogger has posted anything new
- Figure out a nice gift for my niece even though I have no idea what 15-year-olds want these days
Even the last one, where the user can’t describe the outcome, is not about the site. It’s about their niece’s gift. That user (like every other user) would want to surface all the content related to their goal, but will show no interest in content that’s unrelated. Only designers are interested in seeing what’s on a site.
In our studies, almost 94% of quests on web sites have a single objective. When the user reaches the target page, they’ve accomplished their goal. (Or, at least the “finding” portion of the goal. There still may be transactional component, such as purchasing.)
So, in 94% of the tasks, if the user turns to the breadcrumbs, it’s likely because they couldn’t find their target page and are lost. That leaves at most 6% where the user completes their initial objective and needs to start on a subsequent objective: “Ok, I’ve bought the down jacket. Now I’d like a matching hat, scarf, and gloves to complete the outfit.”
Michael’s argument is even if multi-objective quests happen infrequently, the breadcrumbs still serve a useful purpose, revealing the rest of the content to the user:
Assuming you have taken the slightest modicum of care with building bread crumbs, users will recognize them as a reflection of the hierarchy of your site’s information architecture, making them a tool that users have no reason to ignore if they are viewed as an aid to going where they want to go.
But that’s the point: users don’t care about the hierarchy of the site. The thousands of users we’ve observed for the last 12+ years clearly tell us that users don’t care how the site is constructed. Users only care how to get from they page they are current at to the page containing the content they seek. Even with repeated use, they’d prefer that each site visit just have clear scent. Memorizing the nooks and crannies of an information architecture is not their desired outcome.
Michael agrees with this statement from my original article:
We’re recommending that when teams see users needing breadcrumbs, they look for other holistic design solutions. They’ll need to watch users and see the circumstances leading up to how the need arises. In almost all cases, they’ll find a better way to solve the problem than traditional breadcrumbs.
He goes on to say:
The key to successful bread crumbs is that they should be a secondary navigational tool. But, I would argue that people don’t use them because they need them, they use them because they see them as a means to get to where they want to go. As far as the user is concerned, that might be a quick link, an A to Z index, a menu, or a bread crumb (and all of these, minus menus, are generally secondary tools). The thing is most users neither know these terms nor care about them. All they care about is “I click here and go where I want.”
Michael is correct that users don’t distinguish between what he calls secondary navigation and the other types. The idea he proposes, “I click here and go where I want,” is a basic notion behind thescent of information theory: if the target content gives off good scent, users will click on it.
Let’s return to our down-jacket purchaser, now looking for matching accessories. If that user’s trigger words (such as “scarf” or “hat”) appear in the quick link, A-to-Z index, or breadcrumbs (Michael’s secondary navigation tools), then all is well.
Yet, on many sites, it’s dumb luck if the site designers have included the trigger words in those tools. In most cases, the designer hasn’t researched the specific trigger words users will want. Instead, they produce a set of generic terms (“accessories” or “outerwear”, for example) that may or may not resonate with the user.
So, I’d go further to say that all the secondary tools that Michael mentions are also cop-outs: fixing symptoms (in this case, providing a standardized navigation element) instead of the users specific problem (getting match accessories). (I wrote how sitemaps, which are parent to A-to-Z indexes, are also cop-outs in another article.) If I asked any designer worth their weight in salt to design a way for someone who just picked the down jacket to find the desired matching products, I’m betting, of all the design alternatives, Michael’s list would be the last choices.
I agree with Jared that given perfect IA, smart menus, and intelligent visitors, bread crumbs are a waste of time. In reality, few people run sites that function in such a static bubble that one person has control over every facet of how information is disseminated. [...] It’s like saying “In a perfect country, we wouldn’t need laws to punish robbers, because no one would steal from each other.” The reality is, people do steal. That doesn’t mean we shouldn’t strive to stop them, and shouldn’t minimize the problem, but you still must address the issue. So what do we do? We create a ton of secondary navigational elements, build them nicely into our layout, and let the user decide how they want to combine them to go where they need.
In the stealing analogy, it would make sense to look at the economic conditions driving people to stealing. Solve those and the robberies diminish. Focus only on punishment and you end up spending your resources building more prisons indefinitely.
I’d say the same is true for breadcrumbs. Users don’t want choices in their navigational tools. They want clear scent to the content. It’s the designer’s responsibility to provide that. Anything else is just a cop-out.
Breadcrumbs are Simple to Implement
One of the most common objections to my argument is “breadcrumbs are so simple to implement that there is no harm to just doing it.”
Unfortunately, this isn’t true. On a site of any decent size (greater than 500 pages), breadcrumbs become very difficult to implement well.
Often, in an attempt to make life easier, the designers use the category hierarchy as the breadcrumb links. On the surface, this sounds like a good idea. After all, if the categories are well thought out, then they should work in breadcrumbs as well as anywhere else.
Alas, that isn’t the case. Breadcrumbs stand by themselves as solo links. The categories are usually created to be shown as a collection. A category may have a clear meaning when shown alongside its siblings, but is often baffling when shown alone.
Take this example from Michael’s post – the breadcrumbs from NewEgg.com:
It’s not clear what the siblings are. I’m betting most folks would be surprised to find “Networking”, “PCs & Laptops”, and “Apple” to be listed as siblings to “Computer Hardware”, for example. Arriving at links that would describe the entire category well are difficult and usually require more than one or two words. That’s where it becomes difficult to implement breadcrumbs.
Throwing the Baby Out
As I’ve said before, I’m not suggesting that designers stop implementing them. I’m just trying to prevent the knee-jerk reaction of always including them under some misguided notion that they always improve the site.
In the best case scenario, they take no effort (as in automatically compiled by the CMS) and are ignored by users—thus are no harm done. But, that’s rare and unlikely for most situations.
Good design understands why every pixel is in the design. The designer knows how every element is directly serving the user in each instance. Automatic design (“every page needs breadcrumbs at the top, whether we have evidence it helps or not”) rarely accomplishes this.
But, here’s the rub: In the end, it doesn’t matter what I say. It only matters what happens with your users on your site. If Michael’s observations of his users shows that breadcrumbs are the most useful way for them to achieve their objectives, then I think his site should have breadcrumbs—cop-out or not. (And I’d like to learn more about his situation, because I’m always interested in proving my theories wrong.)
Does your site need breadcrumbs? The only way to know is to watch users. It’s simple, really. When we see someone click on one, we stop them and ask what they’re hoping to accomplish. That gives us a use case to work with. If the use cases point to a breadcrumb element being the best solution, then we go ahead and make that work.
Some find my labeling specific elements (like breadcrumbs) as cop-outs is harsh. But, that’s the point. Had I said, “breadcrumbs might not help as much as you think”, you probably wouldn’t have given this topic as much thought.
My purpose is to get you to think twice about using them. If I’ve made you seriously question your usage of them, then I’ll sleep well.Tweet