Rating averaging methods _should_ treat scores with fewer data points as less trustworthy and either suppress showing the score or apply some early-rating bias. I.e. if users are sorting by rating new products should never be near the top.
Otherwise it should be possible to sort products or even brands/sellers by age and prefer older ones with more reviews.
I'm not sure Amazon does the first though ATM, and it definitely doesn't do the latter.
