Correlations are easy to find, and many mean nothing. Causation is a much harder fact to establish.
A recent report claimed that people who take a multivitamin regularly are 30% less likely to suffer from heart disease. The knee jerk reaction is to go out and buy some multivitamins.
But a more considered approach might be to analyse the lifestyles of the sample used for this research. It seems fairly likely that people who take multivitamins are concerned for their health, and so do other things as well. A reasonable diet, some exercise and various other health promoting choices seem more likely with people who take supplements.
This is a good example of jumping to the conclusion that correlation (multivitamins and heart disease in this case) means causation – multivitamins reduce heart disease. The more likely scenario is that there are some lurking variables – variables which cannot be seen, but which account for the link between correlated variables.
This is a tricky subject, and in reality we often do not know if there are lurking variables, and as such conclusions asserting causation are often wrong. With the proliferation of data mining, data visualisation and other technologies in business, we can be pretty sure that people are jumping to conclusions – many of which will be wrong. There is no simple answer, but a good understanding of the domain should eliminate many correlations which are misinterpreted as causation.