Recommended 2-page article in Harvard Business Review, May 04
Management Lessons From Mars: "NASA's fabled Faster, Better, Cheaper initiative sped up the agency's spacecraft development. But when missions began to fail, it was faulty organizational learning--not hardware--that was to blame. "

NASA demanded faster development by increasing re-use, but didn't put the KM in place to make this happen. They compressed the pipeline so much, that new projects couldn't learn what they needed to from exisitng projects becasue they weren't complete yet: "in short, NASA was raising the bar before seeing if project managers could clear it where it was".

Where I'm not sure I agree with MacCormack is his recommendation to "capture all the important learning" and "Institutionalize postmortems on all projects". If you're like NASA where every project has a high degree of exploration and uniqueness, then this may make sense, but if your business involves many projects that are just variations on a theme then you have to ask what the incremental value of each one will be. Moreover, for any sufficiently complex project, its unlikely that learnings can be transferred to people unfamiliar with that context, in which case the postmortem should focus on helping project members structure what they've learned for their own benefit (building organizational competence) rather than capture and transfer of 'learnings' as an object.