In Windows XP, we have 3 tasks that launch wizards where the first step is to confirm the files you want to mess with. These are Order Prints Online, Print this Picture, and Publish to the Web. (Technically, the Image Acquisition Wizard does this as well, but at that point you didn't see those pictures in a folder already).

Why do we do that? The other option at the time was to simply make sure you selected all your pictures and then run the wizards. The problem was two-fold: first, many users have problem with multi-select (particularly when in a "double-click to activate" world), and secondly, the flow was not something everyone expected.

The first problem is clearly important in the XP case, where we focused intensely on making basic digital media scenarios work really well for everyone.  We knew that many of the "moms" out there would have difficulty being able to select a number of photos and print them - making much of the experience painful. For example, we would sit in the usability lab and watch someone run the wizard 10 times over to get every picture and print it out. Ouch! They were technically successful, but clearly as an experience it failed badly (taking often 10 minutes or more to get it to work). The flipside was, of course, that they remembered it - having had to do repetitions. (Perhaps not an ideal way for you to educate users).

So, we decided that adding a confirmation step, optimized around multiple select (note the checkboxes on the items) would work. The checkboxes are hot and users could click on them (and would gravitate towards them) plus by using the thumbnails we could present bigger targets (leveraging Fitt's Law).

The second issue, on flow, may be non-obvious. But the deal is basically that when people are reading left-to-right, the task comes first before the items. The flow can imply that you hit the task first. The other direction is to select items then go back to the left and hit the task - something experienced users easily figure out for the most part, and novices or simply less experience Explorer users don't necessarily discover. This allowed us to once again catch those people.

The design has a couple basic niceties to it: it stays mostly out of the way of the experienced user - you don't have to do anything on this page if you already selected the items you want - just hit Next, it makes it much easier for users to select, and in the case of things like Order Prints Online, where you will be spending money to order prints, it gives you that extra level of confirmation that hey, these are the pictures you will be ordering. This is particularly important for user trust - after all, with that feature we're asking you to give someone your credit card online, and providing that capability to novices and perhaps Internet-beginners can be a stressful experience.

That’s not to say it’s the be-all end-all or there aren’t obvious areas for improvement (there are). But it worked well enough that users weren’t bothered by it and significantly decreased the time and frustration it took to be successful using it.

Multi-select is in some ways a holy grail of design here - when you build an interface to select and activate items, and want to be able to select multiple ones, how do you do it? Windows, in single-click mode, will do "hover select," but that often can annoy power users (or others quick with the mouse). Drag-select is very non-obvious to users. And so on. This doesn't mean the problem can't be solved, but it is an interesting question for balancing of design and flow priorities in a user experience. What we've done here is to place multi-select solutions in context of activity taking place - and only there. You still can do the power user thing, and we try and catch those unaware when appropriate.