Phil Windley posted a synopsis of a $1B enterprise-wide IT project at McDonalds that just got scrubbed. The plan was to upgrade their infrastructure so that everything could be monitored near-realtime. Currently, the aging mainframe-based system can deliver some numbers one week old at best.
Phil raises a few good points on why this project was doomed to failure: the scale was just too damn big, and the project was a Big Up-front design, not iterative. Phil also points out that McDonalds failures don't necessarily apply uniformly to other business, simply because there are so few business with comparable problems and comparable scale.
