14 April 2012

Cobb's Paradox

"We know why projects fail, we know how to prevent their failure -- so why do they still fail?" Martin Cobb, CIO, Treasury Board Secretariat, 1995

Seventeen years - and still no answer.

Actually, there is an answer, but the question seems to have become unpopular.


  1. Hi Marc,

    Cobb is wrong; we don't know how to prevent project failure. What we can do is define what failure is for a particular project and then predict the probability of failure given a set of known finite constraints.

    And you're right, asking these questions is not very popular.


  2. Hi George,

    We can't prevent every project failure, but do we know enough to prevent most of them if we use what we know?

  3. What does "most" mean?

    Let's say we are perfect at estimating the probability of success for projects; then projects with an estimated 75% probability of success are going to fail 25% of the time, and we can't prevent that or predict which of the individual projects will ultimately fail.

    So, if you can estimate the probability of success for a universe of projects and your estimates correctly predict the success/failure ratio of that universe of projects over time, then you have, essentially, prevented project failure to the maximum extent possible.

    But you still have failed projects.

  4. True enough. The best we can do is make the 75% prediction right 75% of the time.

    Unfortunately, that doesn't make a very good pitch.

  5. Isn't there an analogy here with safety ? "We know why accidents happen, we know how to prevent accidents, so why are there still accidents ?"

    Not being able to prevent ALL accidents doesn't mean we shouldn't aim to prevent the preventable ones - and doesn't the same apply to projects ? We may not be able to prevent every project from failing, but should we avoid the steps to make it less likely ?

    1. Project failures are rarely accidental, though it may feel like that sometimes. Of course you're right; we should be taking steps to avoid the known causes of failure.

      One of those steps is to stop using PERT and CPM because they both are fatally flawed. The math they use is wrong and produces consistently optimistic estimates. That's why I've become a probability Management missionary.