Users of E-commerce websites like Amazon, eBay, etc. benefit from a large pool of information available through observing the other users’ reviews and comments. However, this information can be in error, owing to many reasons, one being malicious attempts to enter fake reviews and another being occasional users’ mistakes. In this paper, we model the possibility of  informational cascades (or herding) with erroneous feedback modeled using a binary symmetric channel. Using Markov chains, we formulate the welfare of each user as a function of her signal quality and the crossover probability in the channel. Our main results are that a lower error level does not always lead to a higher welfare and it maybe advantageous for a social planner to artificially add noise to increase welfare.