Usability has always been a big—and sometimes badly handled—issue in information systems design. Ever since Apple launched the current generation of GUI design metaphors (borrowing heavily from Xerox PARC, of course) and Microsoft followed with the Windows series of evolutions, we have lived with usability within a defined context.
Because these (distinct but internally consistent) UIs always behave in predictable ways, they are easy to learn and pretty efficient—but neither is really “intuitive.” The fact that they are almost ubiquitous (even competitors like KDE and Gnome are really just behavioral clones) has been both a blessing and a curse. A blessing because whenever you switch on a system, you know how it works, and a curse because there are many things you can only do the way the UI allows—and customization to suit personal preferences and cognitive abilities is complex.
Since the development and wide availability of the so-called “WIMP UI” metaphor—that’s Windows, Icons, Menus, Pointers—there has been a continuing debate amongst usability researchers on the advantages (common operating behaviors) and disadvantages (enforcing the same behaviors for everyone) of the approach.
We have probably come as far as we can with productivity using so uniform an interaction metaphor. To get the next level of capability we need to add the advantages of personalization (tailoring behavior to the unique capabilities of the individual) without losing the advantages (break fix costs, upgrade-ability, easy management, provisioning of a common infrastructure, initial familiarity) of a common interface design.
All that could be about to change, and it could have a big impact on usability and on how we think about using technology to improve task performance. The new Microsoft Vista operating system, due out later this year, includes the first full makeover of the UI in over a decade. There is no guarantee that all of the possibilities will finally make it into the released product, but if they do—and the current preview releases and associated development kits look promising—application developers will have some powerful and relatively simple to use new tools to “tweak” usability at the level of the individual user.
[Note to the reader: I do consulting work for Microsoft and I’ve been a beta tester of Longhorn/Vista since the very early milestone builds in 2002]
Applying a general productivity principle—make the platform do the work—has the potential to redefine a lot of current unproductive UI behaviors. In particular, the effort required to categorize, store and retrieve information of all kinds now accounts for a significant percentage of knowledge worker time—and the results are at best unsatisfactory for personal productivity and a major barrier to the easy sharing of content with a work group or wider community. My early experimentation indicates that the Vista UI may reduce these efforts on an individual basis by about 30 minutes a day.
The level of group effectiveness improvements is harder to assess. I estimate that small workgroups create three to five times as much content as they “need”—because they can’t easily find what already exists or see what other workgroup members are working on. Not all of this effort can be eliminated—sometimes starting from scratch is more effective than reusing and modifying existing work—but I estimate that at least half of the redundant effort could be eliminated.
In my models, knowledge workers, depending on their work roles, spend on average around half of their time on small-group activities, and about 40% of that is spent on some form of content creation or review. Removing half of this effort would save another 30 to 45 minutes a day.
Secondly, I have been looking at how information presentation options in Vista improve the comprehension and accuracy of response of information workers. Early results clearly indicate that the potential is there for general improvement for all information workers, stemming from the ability to tell what is critical from what is not in a “noisy” information environment and to prioritize accordingly.
But more work will be required to fully understand how to leverage this capability. The problem is inherently about understanding how people communicate effectively—and these kinds of interaction design skills are new to the majority of application designers.
Finally I have started to look at how applications designed to run on, and seamlessly integrate with, the Vista platform might be able to deliver additional productivity through better automation of information worker tasks. There are clear opportunities in workflow and in analytic processes, but once again more work is required.
Despite the many intangible aspects to a post-Vista world, I have attempted to estimate what the aggregate productivity impact of Vista could be for information workers who use a computer most of the day.
My first pass at a model attempted simply to estimate the time saved for each PC-centric information worker from Vista UI improvements, and I estimate that the savings are in the range of 45 to 90 minutes in a working day on a like-for-like workload.
If we factor in the additional improvements that come from the redesign of work to really leverage the capabilities of Vista (and eventually Office System 12), the potential savings approximately double.
Finally, add in the potential productivity from business applications designed specifically for Vista, and the final productivity improvements could be substantial—in the range of 20% to 40% depending on the information worker’s role and responsibilities.
These models need a lot more work, and I have yet to fully assess the cost of developing and deploying the capabilities that make the productivity possible.
With much more powerful user interaction tools and capabilities available, we will need smarter set-up wizards that subtly incorporate user capability tests; we will also need “observed behavior” tools linked to an adaptive model for functionality; maybe using a “simulated annealing” approach (which measures the “length” of the route you use to get to whatever you want to do next, and then moves things around so that the path is as short as possible—think of it as always having what you need just one mouse click or voice command away from wherever you are) to optimizing the user experience.
We will have to think more about the impact on client systems that are routinely used by more than one person. Do we impose a common (lowest common denominator) personality in the client? Do we require explicit user switching when a new user appears? Can we detect this automatically and generate a request for a new user ID? What’s the overhead for this in terms of time and local resources to store different profiles? What happens when we have several people looking at the same instance of the UI? What’s the impact on remote support? Can we look at using smart cards to key the client to the “next” user?
We also need to look harder at an effective taxonomy for types of work amongst “information” workers.
My work has identified several broad categories (based on task repetitiveness, task complexity and duration, range of required tasks, interrupt level and so on), but it is by no means exhaustive, and I have not yet thoroughly reviewed the extensive research literature on this subject. And well-thought-out role-based design templates will be needed for the common types of work and levels of capability—we need to avoid really bad DIY User Experience designs.
Personalization of this type is hard to do as an abstract process—and too much work for most people to do—even once. To be effective, personalization probably has to be progressive, in the background and automatic.
We will need a lot of re-training for information workers to teach them the new habits of the Vista UI and to eliminate the old habits of Windows. We know from the XP deployment experience that old habits are hard to change—more than half of XP users prefer the “Windows Classic” environment because it’s familiar, even though there are features and capabilities in the XP environment that boost usability and productivity.
We probably also need to develop a lot more templates for the organization and storage of everyday information types that are specific to a business or industry context rather than just make do with the new Vista types that are generic to business activities across industries.
We need to think about an efficient way to do this and to protect the IP these type definitions and schemas will represent. DRM clearly has a role here—but probably more important is to focus on a practical pricing and licensing model.
If you’re a CIO or application manager faced with Vista, you have both a challenge and a great opportunity. You will for sure need a transition strategy that’s more sophisticated than the usual upgrade. You’ll need to start by identifying and defining appropriate generic and specialized work types, roles and looking at the associated capability sets.
You should be thinking about developing relevant “everyday” information type templates for your industry (probably as a first step to developing more comprehensive standardized XML vocabularies) and thinking about the extent to which these should be broader standards.
You should start to identify and develop templates for role and task-related notifications and alerts so that messages have appropriate metadata templates for visual and aural volume, placement, duration, repetition and so on.
That’s a lot of new work and new skills—so you probably want to start to practice as soon as possible.
All this raises another interesting point, which I intend to get to soon: Is the added productivity going to be worth the effort of transition? More on this in another piece coming soon.