I really love this kind of report. They are such a good example of the kind of bullshit that you can find in the marketing world. The report draw conclusion about completly build numbers that doesn't explain at all why the TCD is greater in one place than the other. In my opinion, if you can't give the reason, is clearly because your study is incomplete, especially on such a complex case like this.
I don't want to go in deep explanation, [some] Windows Embedded vs Linux Embedded TCD analysis [people] Register Windows Devices Development have done it better than me. However, I want to talk about my specific experience. I have work on Linux Embedded, Windows XP Embedded and Windows CE project. The most innovative ones were always the Linux ones. They involved very small and limited hardware platform (to reduce cost), which were often completly custom. The interface was fairly innovative, even if some ISV solutions were available, my clients/employers always want to have a new, custom and innovative interface, which just augment the risk.
On the other side, all my Windows Embedded projects, which are often a follow-up on previous Linux experiment, was on fairly big barebone PC or other popular platform, already tested and manufactured, and display some interface barely custom buy from an ISV vendor. Functionnalities was simply dropped if the vendor or the OS doesn't support it, bugs was forgiven if coming from the bought parts and innovation was simply inexistent. Briefly, the Windows devices simply look like all Windows devices and can be copied from scratch and easily enhanced by any hacker in his basement in the space of a weekend.
So, is my Linux Embedded projects have greater TCD? For sure, yes, but not because the use of Linux. The risk was higher, the hardware platform less mature or not mature at all, and mostly everything has to be made nearly from scratch by design. Even trying to do the same with Windows Embedded will be simply more costly, if not completly impossible because of licensing, limitation of the OS platform or support issues. Creating the same kind of devices that I do with Windows Embedded with Linux was simply dismissed as ininteresting: who will want to buy a device that any hacker can create in his basement simply by downloading is favorite distribution and software into commodity hardware? Well, that's exactly what happen in the Windows Embedded market. The Windows Media Center can only be buy by OEM simply to create an artificial demand, when in facts it can be installed on mostly any PC with the right hardware configuration. You can even create your own Media Center PC on Windows with simple free software and the right hardware. Is the Freedom to Innovate only available for Microsoft?
So, the lower TCD of Windows Embedded is simply artificial. If we can calculate the ratio between Innovation or Design Complexity vs TCD, the Windows projects will just look very poorly. Is no doubt that was one of the thing not taken into account by the study.