If you survey 10,000 of your customers by email and 200 reply, what will you learn from the responses?
You will probably not get a statistically accurate presentation of how your customers feel. What you will get is an accurate understanding of how customers who answer email surveys feel. Two different things.
People who vote are not always the same as people who answer surveys. People who post Yelp reviews are not the same as people who buy from you. Customers who complain are not the same as all customers.
Sure, sometimes the groups are similar enough that it's okay to use one as a proxy for the other. But often, that's just not the case, and we mistake proximity and noisiness for accuracy.