It never ceases to amaze me.  In windows XP, it was a common joke that Microsoft couldn’t predict how long something might take.  You’d get “4 minutes, 17 minutes, no wait, 2 minutes, no… 2 days 14 hours, no 30 seconds…”  Their estimates were so wildly off that there was absolutely no point in paying attention to them, unless you wanted a chuckle.

In windows 7, I thought they had addressed this issue, but apparently not.  Today, I got this on my screen.  I’m copying two folders simultaneously to a USB external hard disk.  Look closely at the times:

Microsoft's complete inability to predict task duration

So from this, I am to believe that it’ll take an hour to copy 2GB of data, and 6 hours 30 minutes to copy 193G.  That doesn’t seem very consistent.  A few minutes prior, the top one was indicating 23 hours to do the copy, and the bottom one was about the same.  The top one actually took 11 minutes to complete.

Come ON, guys!  This isn’t rocket science.  It’s a very VERY simple thing to sample how long it’s taking  you to do a task, and knowing how much more you have to do, it’s TRIVIAL to calculate how much time that will take.  Sure, you can make it more clever by doing moving averages and so forth to allow for fluctuations in resource availability or WAN vagarities.  Still, it’s NOT difficult to be somewhat close even if you don’t try to refine estimates in this manner.  Other software manufacturers are able to surmount this seemingly impossible task, as their progress indicators actually seem to be reasonably accurate.

I think that at this point, given years of demonstrated inability to accomplish this absolutely overwhelming programming task, Microsoft should simply remove the time estimates from their products.  They are not useful to the customer, and they are embarrassing for Microsoft.  Isn’t anybody over there even slightly embarrassed?  If it were my company, this would get fixed, or heads would roll.  I know there are a lot of smart people at Microsoft – give the problem to one of them instead of the morons who have been working on it to date.

Support for computers and software is notorious for being frustrating and ultimately not helpful.  Unfortunately, this isn’t just a stereotype, it’s a truism.  The two big problem areas I see in this realm are:

1. Outsourcing – many companies outsource their support to cheap labor abroad.  Customers have to wade through knee-deep accents to try to get the person to understand the problem.  Typically, these people do not understand very much at all, but are reading from a knowledge-base “script” as they ask the pre-canned questions and have the user do things to try to solve the problem.
2. Boilerplates – even domestic tech support suffers a similar problem.  There are stock boilerplate responses to incoming queries which are dispatched as quickly as possible in order to move on to the next problem.  Typically, support techs are rewarded for the number of cases they can resolve.  So, if they can get the customer to just give up and go away, the case is resolved, and this looks good for them.  Very little thought is put into the process, and frequently only one part of a trouble ticket will be addressed by the boilerplate response, and everything else is ignored.

The industry clearly needs to fix this as long-term, this could be very destructive.  It’s somewhat like loaning money to people who can’t afford to pay it back… does that ring a bell?