A good survey when carried out randomly should not need that much number of players. 100 players is a far-fetched amount for MD's population that are online. Both the answer and the problem of a survey lies with the margin of error. A low number of responses from players means a higher margin of error, and a high number of responses from players means a lower margin of error. Based on the population of MD that are online, which is usually between 250 to 300 players. Out of a pool of let's say 300, an acceptable amount to sample may be one tenth of the pool, which average around 30 players. After all, most surveyers only select a random amount of 1000 people out of a population of 100000 people. The margin of error is fairly acceptable for an amount of 30 sample from a pool of 300 players.
But certainly unresponse sample may create a problem for the survey as well as stated by Kafuuka. But hopefully the time alloted (possibly a month the same time for MDNP) may compensate for this bias method.
(Not done writing going to edit later on)