Did you have very specific hypotheses to test before doing analysis of the dump? Sentences like "Websites with feature x rank y percent better" are pretty meani

Did you have very specific hypotheses to test before doing analysis of the dump? Sentences like "Websites with feature x rank y percent better" are pretty meaningless if they're just patterns that appeared from the dump. Any large enough sample is going to have significant looking patterns which can have a likely reason back fitted to them, but which are simply due to randomness.