I’ve been blessed to witness at least three “one-in-a hundred” events, in just the last few years. And it has prompted my question: How?
Forgetting the potential for hyperbolae in the media, I am talking about quoted experts making these statements.

As far as I can see, there are three possibilities:

1. The statement “one-in-a-hundred” implies an analysis, over a model, so we can’t blame poor modelling. But poor analysis might be an issue. Are we using small-tailed distributions, such as the “normal” bell curve too readily? Should we be employing a better choice of analysing framework, such as Poisson, Weibull/Gumble or other exponential families?
2. Are the data we are using biased, bent or in some way corrupted? Perhaps we are simply measuring the wrong thing, and applying that observation inappropriately. In what is left of the financial sector, there is no denying this could be the case. But we’ve also witnessed large scale events in other sectors.
3. An alternate is that the underlying systems are evolving rapidly, much faster than our analyses are accounting for. This would imply that our time series are too short, that our data is not aged appropriately, and that the statistical analyses, being necessarily historically based, will always lag.

Whereas the first two issues are fixed by using more appropriate people to perform a task, the last issue offers a direct challenge to some business applications of evidence based management. How do we collect evidence in an environment with a change time similar to our collecion time?

The answer is: Use a different form of evidence - Logical structures from cognitive applications – straight talking, straight thinking.

We can still build an evidence base within this environment. It just looks different as it involves coherent structures of arguments, over the traditional analytical structures. In this application, we build upon asumptions that survive a barrage of logical implications and reductions, as opposed to a data stream that survives a barrage of cross views and tests.

But, if you thought getting a community to accept using evidence based systems over gut feel was hard, and getting emotional arguments out of management decisions was a battle, wait until you try explaining the difference between a logical succession of inference and an emotional subjective opinion that won’t survive cross-examination!!