Why is Windows RT so?

Original author: Hal Berenson
  • Transfer
Немного в ответ топику Microsoft терпит неудачу? решил перевести следующий, на мой взгляд довольно интересный и глубокий пост Is Windows RT the ultimate example of using Telemetry? о причинах, почему Windows 8 и Windows RT получились именно такими. Автор, Hal Berenson, бывший работник Майкрософт, занимавший там достаточно высокую позицию, ныне независимый консультант. Вообще, ведет очень хороший блог, который всем, кто следит за Майкрософт, я советую.

Many puzzle over why Windows RT turned out to be just that? Especially in terms of the lack of support for third-party desktop applications. I believe that Windows RT is the result of the massive use of telemetry in decision making. Those who actively dislike the new Microsoft product can surely recall the proverb that there is “a lie, a big lie and statistics.” But, on the other hand, reliance on statistical analysis can explain why the standard user’s reaction to Windows 8 and Windows RT seems to be significantly better than the reaction of experts and experienced users. It is difficult to positively evaluate something if you are outside the target audience for which the product is designed.



Each decision is based on many assumptions, the most important of which is data. There is always not enough data for a solution (hence the saying that management is the art of making decisions based on insufficient data). Someone might think that with excellent data, decisions are easy to make because the answer becomes obvious. But usually you have to mix data from different sources. The more sources, the more potential errors may appear as a result of the analysis process. Often these sources themselves are the result of the analysis of the primary data, in which there were errors. And, in the end, you can get solutions that have nothing to do with the reality of life (for example, New Coke ).

But what if you have nearly perfect data? What if, instead of small volumes, limited collection technologies, random data, there is a huge sample of which we can say that it absolutely represents reality? Could this lead to better solutions? Especially when these solutions are of high complexity and high risk? It seems that a big experiment with Windows 8 will eventually answer this question.

When Windows Phone 7 was being designed, the development team investigated the 100 most popular iPhone apps to make sure the platform could realize all the features it needed. The basis of the decision was the study of use cases, but this is not direct data. The Windows 8 team, in addition, had access to a huge amount of data collected using the CEIP program from those users who voluntarily signed up for it ( or did not unsubscribe when using beta versions, where it is enabled by default - approx. Per. ).

Anyone who follows the Building Windows 8 blog could notice how seriously Microsoft took these data when developing the new Windows 8 interface. Removing the Start button (to make desktop users jump to the new start screen) is just an example of using telemetry about standard system usage patterns. Of course, if you are in the minority of those whose usage pattern is very different from the standard, you will not be happy with what Microsoft did. And even, just that Microsoft has data that talks about how people really use their product does not guarantee that the solution will be correct. I cannot tell the data, for example, that the user wants just such a switch between the old and the new interface.

But back to Windows RT and how telemetry could be used to make key decisions. Recall netbooks. Five years ago, it came to be understood that the use of computers is moving to the network, and that users need an inexpensive special device for this. Such a device would not be a full-fledged PC replacement for most users, but was considered as an addition to the main computer. Initially, netbooks were not intended to run existing windows applications, and therefore manufacturers focused on Linux, as the main OS for them. The rapid growth of this market made Microsoft pay attention to it and offer first an inexpensive version of Windows XP, and then Windows 7 Starter. Despite an additional 15% of the cost (partly due to the cost of the license,

The netbook market share skyrocketed, became a quite noticeable niche in the PC market, and after that it received a triple blow right away. In 2009, the App Store was introduced and began to grow rapidly, providing an alternative to leaving all applications on the web (plus the iPhone had a pretty decent browser). Then the demand for netbooks shifted towards light and thin 11-inch full-fledged laptops. And finally, Apple introduced the iPad, which was a much better alternative to a netbook for using the web, and also had the ability to run applications from the App Store. Netbooks have almost disappeared.

Windows 8 development began even before the introduction of the iPad. Even later, during the launch of the iPad 2, many analysts considered the tablets to be nothing more than a replacement for netbooks. And until recently, the impact of tablets on PC sales was only in the sense of a shift in demand for them from netbooks. So, in terms of the availability of data for making decisions about the design of Windows 8 and Windows RT at the end of 2009 and beginning of 2010 - telemetry from netbooks through CEIP is what Microsoft had at its disposal.

Now we focus purely on Windows RT. In 2009, it was already clear that a huge ecosystem was growing around ARM processors. Microsoft has followed, and even worked on porting Windows to ARM since the beginning of the decade. The decision to port to ARM is an easy part of the task. The decision about what to do next with him came through telemetry analysis. And I’ll be clear, I don’t have information that it was really that way, but I’m ready to put the money, that it is not far from the truth.

Since the Windows Phone team was in charge of the phones, and ARM definitely could not carry out the heavy tasks for which they use full-fledged laptops and desktops, it was clear that Windows on ARM should focus on the class of devices between the two. At that time, only one device was there - this is a netbook. Once again, the iPad did not yet exist, and the reaction of users to it was not known. But the general characteristics of tablets similar to the iPhone then clearly could already be foreseen.

So it was obvious that Windows on ARM should aim at the niche of netbooks, netbooks with touch support, and tablets with features similar to netbooks. But how, in the absence of knowledge about what happened later (the disappearance of netbooks and the rise of tablets), could Microsoft make decisions? Telemetry.

Why did 90% + percent of users choose to pay more for a netbook with Windows instead of Linux? If the device is used only for web browsing, such user behavior is meaningless. Of course, we can only speculate further. The habit of UI, compatibility with a large number of printers, the ability to run regular windows applications (although this contradicts the original idea that was behind netbooks), etc. As I said, we can only speculate. And analysts can interrogate users and draw their own conclusions. But Microsoft? Microsoft has accurate data from CEIP.

Microsoft could take the data and see how often users print and which printers they use. How often do they use USB, and what do they do with it. How often a netbook is used with external monitors, keyboards and mice. How often is WiFi, wired ethernet and 3G used. Microsoft can see how often the user uses the browser, and what types of sites they visit. What other applications does it launch, and how much time does it spend inside them.

And what do you think exactly got Microsoft from telemetry? I guess they saw what netbooks mostly use for the web. Then comes the small but notable part of using Microsoft Office. And then a complete fall in the use of any other applications. Netbooks are basically a device for the web and office applications. Then they looked at the statistics of site visits, and saw that most of them fall under the concept of applications for the consumption of content that have become so popular on the iPhone, and for which a new Metro application model should be developed. And they saw the massive use of standard Windows functionality: support for various peripherals, working with a network, etc. Now combine this data from the use of netbooks with the development of the tach and boom of applications in the App Store, and you will get Windows RT.

The standard question is, why not make Windows RT support for legacy x86 applications through emulation? At first glance, this is not technically impossible. DEC Alpha processors ran x86 applications through emulation, but remember, Alpha in those days was faster than x86 itself, which allowed for sufficient performance. Any current ARM processor is noticeably slower than the equivalent x86 (despite being much more energy efficient). So x86 emulation on ARM will make most applications unsuitable for normal use. But, more importantly, if the data on the use of netbooks shows that the user still does not start anything even when there is such an opportunity, why bother doing this?

Well, emulation is difficult, but why not give third-party application developers tools for porting to ARM and allow third-party applications to be installed there? Of course, there are problems associated with energy consumption, memory usage, and security. etc., from which classic applications suffer, and which just had to solve the new API. Microsoft could force third-party developers to pay due attention to all this, as it did, for example, with the Microsoft Office team. But, this would distract them from creating applications for the new API. And why bother with this at all if users are not really going to run them on this type of device?

It is unlikely that many users would run Photoshop on netbooks. If they used netbooks for photography, most likely they used lightweight applications that filled the App Store and that would appear quickly enough in their own Microsoft store. So, the analysis of telemetry data on the use of netbooks led them to the conclusion that in reality only a small part of users needs to run desktop applications on devices of this class.

Now let's look at Windows RT, or even better at Surface, and see what it is. Surface is where the netbook meets the iPad. It transfers exactly what most users loved about Windows on netbooks in the modern era, removing things that netbook users simply didn't use. This is exactly what users told Microsoft through telemetry data extrapolated from those long-standing netbooks into the modern world.

Another evidence? Domains On the one hand, it seems strange that you cannot enter Windows RT into a domain, on the other hand, how many netbooks were in the domains? Microsoft could have had quite a few other reasons not to do this functionality, but the main factor was that telemetry analysis showed that it was not important for this type of device. Now, recall any question you have about the functionality of Windows RT, and most likely you will find your answer in typical scenarios for using netbooks.

Using telemetry may explain why Windows 8, Windows RT, and Surface
better perceived by ordinary users than by experts and experienced users, whose use cases deviate greatly from the average. Windows 8, Windows RT, and Surface were designed based on current usage data for a class of devices that are usually ignored by experts and experienced users. This segment occupied about 20% by volume before the post-PC era began. If Microsoft really wisely used the full amount of telemetry to do something that made netbooks so popular, plus it added the ability to at least partially cover the scenarios for which Apple created the iPad, Windows RT could actually win.

What if Windows RT fails? This may be, for example, because the discontented voice of experts will clog the product before it gets a real chance. Or as a result of poor decisions that were made despite excellent data. Or because of serious mistakes in marketing, sales, errors of partners who are not related to the product, as such. Or because it will become a victim of a lie called statistics.

Also popular now: