Tuesday, October 12, 2010

Decoupling for threats

Here's another musing on SWOT. If you missed my earlier missive, click here.

"Threats", as given in most explanations of SWOT, are future, probabilistic, external events or conditions--in other words, risks--that could stall, cancel, or redirect project activities.

And, given this explanation, what remedy is available internally to the project? [We'll stimulate to situational awareness that should always be in effect, and could provide an opportunity to work mitigations in the external community]

The best remedy comes right out of system engineering: loosen the coupling among WBS elements, especially final deliverables, as much as possible. That way, if lightening strikes one aspect of the project, the other workpackages continue onward.

If you are a sailboat guy, it means every sail should have rip-stop seams that 'decouple' one segment of the sail from another. If one segment blows out, the whole sail is not lost if there are rip-stop seams.

There are other examples: firewalls, and diversified investments to name two more

So, how do you know if you've got loose coupling in the WBS, and how would you measure it? The system engineering measures are sensitivity and correlation. And for risk managers, there are measures of diversification, statistical independence, lack of a Bayes effect between outputs.

Sensitivity is a measure of the intensity of impact--or intensity of reaction--on one element caused by another. High sensitivity means a strong reaction to a stimulus.

Correlation is a measure of the coupling, or transmittance, of the impact from one element to another--in other words, it's the 'resistance' of one element to another. Low correlation means a high resistance to the transmittance of effects from one to another WBS element.

Diversification is a practice that means: "don't put all your eggs in one basket....have multiple baskets that are independent of each other". One basket gets damaged, the others don't. For risk managers, this is detectable in a Monte Carlo simulation: if the variance--meaning the statistic called 'variance'--of the output seems to go up and down with the number of elements in the simulation, then there is good diversification and statistical independence.

We explained Bayes in other musings. It simply means that if one project probabilistic phenomenom affects the probability of another, there is a Bayes coupling of probabilities from one to the other. There are ways to measure the Bayes affects, and most risk managers are aware of how to do it.

Have someone check out your WBS: you might be able to avoid a lightening strike!

 Bookmark this on Delicious
Are you on LinkedIn?    Share this article with your network by clicking on the link.