The reasons why Big Data projects fail isn't a fault of 'Big Data'
The reasons Big Data projects fail is a complex one.
1) Management believed the hype and their expectations were not met.
2) Those who can work well in the 'Big Data' space are very senior people who have work experience in distributed systems and projects. Techniques that worked well in traditional software development don't always translate to things working well in distributed systems.
3) To be good, you need to spend a bit of time looking at a blank wall.
That means you need to spend time thinking through your problem set before writing a lick of code.
4) Teams need to experiment more and more time doing R&D.
5) Big firms like IBM, Accenture, Deloitte don't understand that you can't take a 10 week wonder and then put them through the basic Hadoop week long training and expect them to perform.
6) The 'hack' (cough) mentality of FB doesn't work well in the real world.
7) Solutions architects need to spend more time thinking about a problem...
8) Developers have to be flexible.
Note while I've said this twice (#3 and #7) its that important.
Posted anon because I do 'Big Data' problems for a living ... ;-)