Besides the other mentioned issues, using average user rating as a proxy of quality is something that needs at least an argument why you expect the ratings to be comparable. Consider that for example shops near a train station will see a lot more one-off customers who might be inclined to rate differently compared to the shop's neighbors.