Engineering Windows 7

Welcome to our blog dedicated to the engineering of Microsoft Windows 7

Designing Aero Snap

Designing Aero Snap

  • Comments 76

Here’s a behind the scenes look at the design of the Aero Snap feature in Windows 7.  We thought it would be fun to take a look at the overall design process of the feature and the tools and techniques used.  This feature poses a unique design challenge in that you just use the feature without any user-interface specifically to invoke it.  As with all features this is a collaboration across all of our engineering disciplines.  For this post, Stephan Hoefnagels, a Senior UX designer, presents the design perspective.  --Steven (P.S., keep an eye out on the Microsoft MIX conference this week!)

In Managing Windows windows and Follow-up: Managing Windows windows we talked about, and you shared, some interesting window management scenarios that we might address in Windows 7. We also touched on some data around typical configurations, as well as goals that guide our thinking in this area.

In this post we’d like to have a closer look at the Aero Snap feature that many of you have already been able to experience in our PDC builds, and of course the Beta. We’ll briefly describe the feature itself, but mostly we’d like to invite you to take a behind-the-scenes peek at our design process so far, and share our iterations, challenges and considerations.

Goals and scenarios

As we explained more in depth in our previous post, our top goal for the Aero Snap feature is to provide you with an effortless way to position your windows the way you want them. We want to reduce the number of clicks and precise movements needed to perform common activities. In a general sense, we want you to be able to manage your windows with confidence and create a feeling of power and control. This is something we touched on in our post on the Windows 7 taskbar as well, and really a theme that weaves through much of our new desktop experience.

Before we look at how we address our goals in the design, a quick note. In the scenarios below you’ll notice sequences of interactions that are fully written out. Sequences like: select the window, click the caption button, and then resize the window. Looking at interactions at this level of detail makes for somewhat awkward reading. These individual steps are so small, and so frequent, that for most of us they normally barely register. Why not simply gloss over some of those details? Well, we spell out sequences of events this way to force us to be consciously aware of the amount of “overhead” that is sometimes involved to get to the task at hand. It forces us to realize what we normally might not. Also, besides providing insight in the problems, it provides us with the right level of detail to consider our solutions.

Now, let’s look at the design!

Side-by-side windows

As many of you mentioned, doing a drag drop operation from one window to another is sometimes a pain. Windows tend to overlap and to get them positioned right can require a lot of fine mouse movement. Oftentimes the steps are as follows: select a window, resize it appropriately, and position it on the screen so that enough of it shows for it to be a meaningful drag or drop target. Then repeat the same actions with the other window. Similarly, comparing content in two windows requires a lot of mouse clicks, resize actions, careful arrangement, window switches, and a fair amount of mouse mileage.

With Aero Snap you can grab a window and move your mouse to the edge of the screen and the window will resize to fill half the screen. Repeat with the other window. Now with two easy motions you have a setup that makes both of these scenarios much easier to accomplish!

Put windows side-by-side with Aero Snap by moving the cursor to the edge of the screen (left to right, top to bottom)

Put windows side-by-side with Aero Snap by moving the cursor to the edge of the screen (left to right, top to bottom)

Vertical maximized windows

We know our users love the maximized window state. Many love it so much that they maximize all of their windows and never even run in any other state! However, with screens increasing in resolution and widescreen layout becoming more prevalent, the maximized window state can lose some of its appeal in certain cases. E-mail is an example. Reading long lines of text across the screen is not ideal. Your eye simply cannot track a line all the way across. Web browsing is another example. Content will sometimes not fill the entire width of the screen, leaving a lot of unused white space on the side.

Now, with Aero Snap you can you can maximize a window in the vertical direction only. When you resize a window to the top of the screen, it will also resize all the way to the bottom. Great for reading long blog posts!

Vertically maximize a window with Aero Snap by resizing the window to the edge of the screen

Vertically maximize a window with Aero Snap by resizing the window to the edge of the screen

Moving maximized windows from screen to screen

We realize there are a few multi-mon users out there, especially amongst the readers of this blog ;-). Ever wanted an easier way to move a maximized window to another screen? A way that’s quicker than clicking the restore caption button, moving the mouse to the title bar, dragging the window over, and finally clicking the maximized caption button again? With Aero Snap you can simply drag a maximized window down, move it over and snap it to the top, all in one gesture. Finally!

Arranging windows on laptops

Arranging windows on a desktop PC can sometimes involve excessive fiddling, fine mouse movements and lots of mouse mileage, as touched on above. On laptops this situation is further exacerbated by the lack of a mouse, making some of these movements even more cumbersome. For these scenarios, and our power users, we’ve introduced some convenient key combinations. Hold down the Windows key and one of your arrow keys to give it a try. You may want to try holding down Shift as well, especially if you’re in a multi-mon setup.

Design process

OK - So those are the scenarios, and the way we chose to address them. Seems straightforward enough, right? Especially for those of you that have been using Aero Snap in out Beta build. It all behaves as you’d expect. But how did we get here? Well, in this next bit we’d like to something that we don’t do that often and really give you a peek into our design process so far and some of the snags we ran into along the way. Keep in mind that this is a brief overview and by no means exhaustive. While we allude to usability testing for instance, this post is by no means meant to give insight into the scope of those efforts. There will certainly be opportunities to talk about some of those topics in the future.

Let’s have a look!

Sketching

We’re back in early 2007, and after we established window management as a potential area of interest, and identified several appealing scenarios (again, see our previous post for more detail on that stage of the process), we started with brainstorms to generate ideas. “How can we make window arranging more efficient?” “More direct?” “More fun?” are some of the questions we asked ourselves. This was a multidisciplinary process, with disciplines like design, user research, program management and development involved. All in all there were around a handful of us, certainly less than a dozen, thinking about this space at the time.

We imposed a significant constraint on ourselves: we wanted to achieve our goals without introducing any extra widgets in our UI. Imagine an extra caption button on the window title bar for instance: this may not seem too bad for one window, but when we’re talking about 10 windows or more, we’ve all of a sudden introduced a significant amount of clutter on your screen. And that’s something that we simply didn’t feel good about. After all, as we shared in our UX Principles talk at the PDC, we believe in “Solving Distractions, not Discoverability”.

We captured our, many, ideas in very quick sketches that we shared via an internal website. Transferred from the whiteboards in our offices and hallways, these took less than 5 minutes to sketch each. The sketches below are some actual examples in which you can start some of the Aero Snap ideas forming.

Early ideas are captured in quick and disposable sketches

Early ideas are captured in quick and disposable sketches

Interactive prototyping in early code

With so many ideas on paper we were eager to try out the best ones. Now was the time to prototype. Note that we are still very early in the process. We wanted our prototypes to be interactive, and we wanted to be able to live with them in our day-to-day work. So we chose to implement the ideas using early code that we could run on our work machines. For example, the image below shows a “smart resizer” prototype running on Windows Vista. Of course these prototypes are not “done” features that we could actually ship: they merely get the basic ideas working, and they definitely have more than a few “quirks” (bugs ;-)). What’s important however, is that they allow us to experience just enough of the interactions ourselves, as well as get feedback in usability studies.

Early “smart resizer” prototype running on Windows Vista, note the taskbar button created by the prototype (and the date in the calendar to get a sense of where we are in the process)

Early “smart resizer” prototype running on Windows Vista, note the taskbar button created by the prototype (and the date in the calendar to get a sense of where we are in the process)

We use this firsthand experience and early usability feedback to iterate on the ideas as we hone in on the final design. We approach this part of the process in a very open minded way and allow ourselves to be surprised. Sometimes ideas that don’t look like much on paper prove to be startlingly powerful. On the other hand, we found out that others look much better on paper than they were in practice! Since we’re not invested in any one idea or implementation yet, we freely refine, and drop ideas.

Finally, the prototypes ease us into thinking about design details. And we might stumble on some insights too (of course we tell ourselves the real feat is to recognize the insight and hold on to it). Here’s an example from an e-mail at the time from one of the team members about a new version of the “smart resizer” prototype:

I noticed something changed. In the original version if I resized it to the max it “supersized” the window. Then if I resized the window smaller, it jumped back to the normal restored state. It was as if the supersize state was different than restored state. With this version when I supersize the window and then resize it again later, it doesn’t jump back to the previous restored size. There was something kind of nice about the supersized state being different than the restored state. We should think about it more and consider making that a part of the design.

After many a prototype we settled on the concept of side-by-side windows and vertical maximized windows. We’re getting clearer on what we want to build.

Detailed design: state transitions

OK - We’ve whittled down our ideas to a good, but not fully specced, set of interactions and behaviors. Time to start filling in the blanks by asking detailed questions. “What does it mean to have a side-by-side window in a world where there used to be only minimized, restored and maximized windows?” “How exactly would you get to and from this new window state?” The e-mail snippet above already pointed to some of these questions.

Currently the common window states are (left to right) minimized, restored, and maximized, how would side-by-side and vertical maximized windows fit in?

Currently the common window states are (left to right) minimized, restored, and maximized, how would side-by-side and vertical maximized windows fit in?

Let’s look at the state problem in detail. Below is an example of two proposals made during this time that show how you can move from one window state to another, for all the different states. Which model is better?

Two proposals detailing the various ways we can transition between states

Two proposals detailing the various ways we can transition between states

To answer that question we considered more specific questions like “What states do we want to link directly and how do we move between them?” “Is it compelling to go from a vertical maximized state directly to a maximized state?” “Should the vertical maximized and side-by-side states behave similar, as they look similar?” Our answers of course guided us to the model that you are now familiar with, which is model B.

But that’s not the end of it. More details need to be worked through, and more questions come up. “What if I want to move a vertical maximized window sideways?” “Resize its width?” “Then pull it down?”

Soon you’ll find yourself in elaborate sequences of events, many possible actions, and even more possible outcomes. Which would be the most expected when actually using the entire system?

To help guide our decision making process we established some guiding statements. Assumptions that we hold to be true. Examples are: “The intuitive way to undo an effect triggered by a mouse movement is to make the opposite mouse movement.” And: “It should always be effortless to go back to the previous “restored” state so as to avoid excessive work to get the window back to a reasonable size.” Or: “If the user specifies a width for a window in a given state, that size should be preserved across state changes when it makes sense.”

Using these statements we were able to answer questions like the ones above in a predictable way and as a result craft a predictable experience. And while the underlying state transitions and rules are fairly complex when added all together, the resulting behavior is, we hope, intuitively understood. That’s definitely something we’re aiming for.

Conflicting rules

Here’s an example of a sequence problem for the really attentive reader. When working through our many new state transitions, sometimes the rules that determine what should happen, conflict. Consider the following scenario. Two rules in our new window model are:

  1. Dragging a window to the top of the screen maximizes it
  2. The intuitive way to undo/cancel an effect triggered by a mouse movement is to make the opposite mouse movement

OK – Let’s try the following scenario in your Windows 7 build. Start with a restored window. Vertically maximize it by resizing it to the top of the screen. Release the mouse. Drag (don’t resize) down the window and drag it to the top again, all in one motion. Release the mouse. What happened?

Your window should be maximized. Which means in this case we chose to follow rule 1. We could have also followed rule 2, in which case your window would be vertically maximized. We figured rule 1 would more accurately reflect the user’s mental model for this scenario.

This is just one example of the decisions we had to make for each transition to and from a window state.

Throughout this process we made sure to preserve the subtleties in the current model. One small example that some of you may be familiar with that we did tweak is the scenario of dragging a window title bar off screen. In previous Windows releases we would snap the window back halfway in an attempt to provide you with just enough space to still move the window around, while optimizing vertical space as much as possible. We chose to be a little more deliberate and straightforward here by snapping back so that the entire title bar is visible. This way you can start to rely on the fact that all of your windows are always very easy to grab, also with touch, and move around. If we really felt half of a title bar is enough, why don’t we always half it, right? For our vertical maximized window states we of course chose to keep the entire title bar visible as well, leading to a solid story all around.

Storyboards and other visualizations

State diagrams are of course only one way to look at the world. We used various ways to communicate different aspects of the feature, picking our medium based on familiarity, availability and intent. We didn��t even shy away from the occasional ASCII art as you can see below! We’ll simply use whatever tool gets the point across. Most of all, interaction storyboards were a very valuable technique to help us understand the user flows, and even though this is only a small sample, you can see we did quite a few of those.

Feature details are communicated throughout using appropriate means

Feature details are communicated throughout using appropriate means

Accidental triggers

Besides figuring out the right state transitions, one or our biggest discussions was around when a state transition should exactly occur, or in other words: when the feature is triggered. We talked a lot about “accidental triggers”, i.e. running into the feature without deliberately setting out to do so.

Aero Snap is triggered by touching the edge of the screen with the mouse

 

Aero Snap is triggered by touching the edge of the screen with the mouse

From the very beginning we always made sure that our feature did not get in the way of current scenarios such as tucking a window off-screen to the side. After all, we don’t want to have a detrimental effect on your current, expected, way of managing your windows. That’s why you have to literally touch the edge of the screen with your mouse, not the window edge, to trigger the transitions.

However, at this time, the feature as you know it now was different in one very important, fundamental way. In our early builds, Aero Snap followed an “instant commit” model. When you moved your mouse to the top of the screen, your window would literally be maximized while you’re still dragging. That is, before you even release the mouse. Moving back in one motion would literally restore the window. We liked this approach as the “preview” was very accurate, i.e. the preview was the window, and there is a certain directness in not having a commit model.

Because our UI is invisible by design, we expected some accidental triggering. In fact in some sense we were relying on accidental triggers to help with the discoverability of the feature. However, after living with the builds for a while we got a little worried because accidental triggering seemed higher than we expected. From our telemetry data we saw people running into the feature, and then cancelling it nearly twice as often as committing. In our usability studies we observed confusion as to what exactly triggered the behavior. Was it the window touching the edge that did this? A gesture? Or something else?

We’re now in early 2008. What should we do? Cut the entire feature? We actually did consider this as an option. Again, we really respect our current window management behaviors, and the last thing we wanted to do is degrade the experience. More constructively, we took on the challenge to come up with a design tweak that would address these issues. We explored several solutions. Some conservatively centered on smoothing out the resize transitions, so an accidental trigger would at least be smooth. A conservative approach for sure, but probably not satisfying enough as a real solution. Others focused on trying to detect user intent more accurately, based on the angle of motion into the screen edge, or the speed of the motion. This proved much too complex to be predictable. Maybe we could trigger the transition only when double-bumping the edge? We were worried that this would degrade the experience of fluidity and flow of the current model.

In the end we settled on the implementation you’re familiar with. We don’t trigger until you release the mouse, making it easy to back out of the effect before anything happens to your window. In addition we provide you with a smooth preview animation, and a cursor effect to help you understand what just happened. This way you can be more deliberate in the future, and use the feature to your full advantage.

Did we solve the issues? Feel free to let us know :-)

Look and feel

It’s interesting to think back and realize that up to this point we had essentially designed a feature without any visible UI whatsoever. Now all of a sudden we have window previews, and a cursor effect. What should those look like, and how should they behave?

Well, luckily we had some things to go on. At this point we had already established a clear picture of our taskbar look and feel (we call it “personality” and will talk about it more in a later post). We had settled on using glass sheets for Aero Peek. We saw an opportunity to use the same visual representation for our preview windows. But how should the glass sheets appear? After experimentation, we settled on a scale animation that originates from the cursor. This gives a subtle hint as to where this preview window came from. We also made sure to animate our transitions. Try this in your build for instance: move a window to the top, and then to the side, in one motion without releasing the mouse. Notice the smooth morph of the preview? Why did we spend time on this? We believe “Small Things Matter, Good and Bad” of course.

Light effects are used to indicate the snap trigger, and glass sheets for snap previews

Light effects are used to indicate the snap trigger, and glass sheets for snap previews

Finally, we tied into some of our ideas around “light” for the trigger indication. We tuned this to be noticeable, but not too loud and made sure to synch with other light effects in the system such as our touch feedback.

We hope this post has given you some insight in the Aero Snap feature, including our design process. We would love to hear your thoughts!

Stephan Hoefnagels

Leave a Comment
  • Please add 6 and 3 and type the answer here:
  • Post
  • Love it! Just 1 more addition... Please can you allow owners of 2560x1600 displays to have extra trigger points in the corners! Could we have it so we can modify it so we get a top right, bottom right, bottom left and top left snap point too by dragging to corners or edges near corners while leaving the middles of the edges for side by side.

Page 6 of 6 (76 items) «23456