I've recently had a look at some Hitwise traffic data for a number of competing retailers. Looking at the graph of the traffic, I wondered if there was any statistical relationship between them; was there any sort of correlation between them? What hypothesis could I form based on that correlation? Here's a sanitised version of the graph. I've added on the Pearson correlation coefficient.
For those unfamiliar with Pearson, it is a way of comparing two independent random variables, it has a value ranging from +1 (roughly speaking, when one variable increases, the other always increases) through to -1 (when one variable increases, the other always decreases). 0 indicates no correlation at all.
In this case, you can see that there is a fairly strong correlation between competitors A & B, but a weak correlation between competitors A & C and B & C.
So what does this tell us? Well, I'm not exactly sure to be honest. One theory is that competitors A & B share traffic, and that traffic would tend to visit both competitors or at least have common visit patterns. It would then follow that competitors A & C (which have nearly no correlation) have a completely different traffic profiles, and potentially very different audiences. So, if I were competitor A, should I be trying harder to capture competitor C's traffic, they are an untapped audience?
Another theory is that competitors A & B share similar marketing strategies/calendars and there is a causal relationship between their actions and traffic levels. But then, you wouldn't expect them to be quite so coordinated.
Another interesting idea is that if marketing activities were actually effective then you might expect there to be a strong negative correlation, i.e. one competitor has a negative effect on the other. However, that's not what the data suggests.
Have you got any thoughts on the subject? Please Tweet me, I'd love to hear them!