Here is a great article discussing the need for better practices in scientific computing:
http://www.americanscientist.org/issues/num2/wheres-the-real-bottleneck-in-scientific-computing/1