I've been doing some evaluation of the networking interfaces on Windows XP and Mac OS X lately. As I've been doing this, I've run into various inconsistency issues. Each time I notice one I find myself asking "Is this important, and if so, why?" I thought I'd share one example and use it to offer some thoughts on when and why consistency is important.
When I started working in the user experience field, I was very concerned about rigorous consistency. I'd often say things like "This needs to be consistent with that, or users will get confused!" I was involved in various efforts at writing UI style guides, and it is easy to approach those with an attitude of "everything must be just like this (or users will be confused)!".
As the years have rolled by, I've watched other folks come into UI design with the same attitude. So, I think it is a natural tendency. Consistency is relatively easy to identify, measure, and enforce. When there is no consistency, the situation is overtly bad: I still remember the chaos in the latter days of DOS. And a cursory use of command line interfaces on any *nix environment will show an amazing lack of consistency.
At the same time, I've also come to see that inconsistency sometimes doesn't affect the user, and sometimes is not even noticed (especially if the interactive flow is right and the inconsistency doesn't require the user to learn anything new) Given limited energy, there are more important things for a designer to do than get absolute consistency (though, your peers and management may be eager to criticize inconsistencies with alarming ease and vehemence).
So, let's talk about one tiny consistency example, and whether it is important.
When should you use an ellipsis (...) on the label of a button? Different platforms have slightly different rules. On Windows, the rule is found in the UI Guidelines:
If a command requires additional information to complete its intended action, and you use a dialog box to supply that information, follow the command with an ellipsis (…). The ellipsis provides a visual cue that the command requires additional information. ... However, not every command that results in the display of a window should include an ellipsis. Do not include an ellipsis for commands that simply display a window or view or change the existing view within a window. Similarly, do not use ellipses for commands that display a collection of objects or options, unless the intended action requires that the user select or confirm the selection of one or more elements of the collection. Also, do not include an ellipsis for commands that may result in a confirming message box
Mac OS X UI Guidelines say something slightly different for that platform, but this example is about Windows, so let's ignore that.
Consider this property window from Windows XP:
Notice the "Install..." button. This doesn't operate on the selected item above it. Instead it displays a window where you can choose something to be installed. Meanwhile, the "Properties" button just displays a window that shows the properties of the selected item and requires no input from the user. This seems consistent with the rules.
Yet, elsewhere in the exact same window, you can find this:
Why does "Settings..." have an ellipsis after it? Isn't that effectively the same as "Properties"?
Oh no! Inconsistency! It is the end of the world!
Well, OK, it is strange that there is this difference in the same window. But, is it really the end of the world?
For average Windows users, I suspect the presence or absence of "..." is not even noticed. It is just visual "noise" that they ignore among the flood of other "noise" in any computer interface. However, the ellipsis is a signal for more knowledgeable users.
So, for the knowledgeable users, is this inconsistency important?
In practice, I think knowledgeable users end up seeing the ellipsis as "Don't worry. If you click this, nothing crucial is going to happen without your having a chance to stop it". So, the presence of the ellipsis provides a sense of a reassurance. If it is used in a case where no action will be performed (e.g. in the "Settings..." case) the central concern of the user ("will something bad happen if I click on this?") is not violated, and so the user doesn't care. (In contrast, if no ellipsis is shown on a button that does display a window before performing some dangerous action ("Format") it may make the user nervous about clicking on it. An ellipsis on a button that doesn't display a window before starting a dangerous action ("Format...") would terribly violate the user's sense of security).
At this point in this example, we can say "Strict consistency isn't important, as long as you don't negatively violate the user's sense of security with the software." As far as the interactive usability of the software goes, that is probably true.
Yet, that statement bothers me. First of all, one of the benefits of consistency is that it allows someone to transfer their knowledge from one part of the interface to another. Put another way, it allows people to work in a new situation without learning something new (and, really, most people hate being asked to learn something new, especially when they are trying to get some task done).
Even if the user doesn't have to learn anything new, I'm uncomfortable saying consistency is not important. When I try to articulate why, I find myself thinking about the issue of "professionalism". When you go to a job interview, you tend to dress nicely. This doesn't mean that you will dress like that every day on the job. Nor does it say much about whether you are actually qualified for the job. Yet, if you don't dress well, it nonetheless reflects badly on you. In many ways, I have come to think this is an important aspect of consistency: If you do the same thing in noticeably different ways in different parts of your interface, it may not actually affect the interactive usability or the learnability of your product at all. But, it may reflect badly on the "professionalism" of your product, which reduces people's trust in it.
To summarize my point here, I'd say: Consistency is important, not necessarily because it improves interactive usability, or because it increases learnability, but because its absence suggestions unprofessionalism. Yet, essentially no users are so detail-oriented to care about small inconsistencies here or there. And the time you spend, as a designer, trying to get every detail perfectly consistent is time that you aren't spending on more important things (e.g., good interactive flow, or a good user model). Let the small details go, both in your designs and when reviewing other people's work.
(as a postscript to this example, I would also like to suggest that the rules of ellipsis use on all platforms are so complex and unnatural that almost no one can get them right. I certainly think that the rules articulated in the style guides are missing the point, that what I think users simply need is a reassurance that when they click on a button or menu item, nothing bad will happen. There may be a much better design solution for that need than putting "..."'s on various controls).