"This is useless data. The only valid data is generally collected by the manufacturers themselves or through some larger, high volume dealers."
Data generally speak for themselves. The sample may be problematic, but unless the respondents are lying, the data (plural, by the way) contain some information. For instance, if 24 respondents report reliability problems, then we are pretty sure that there were at least 24 defective units in the field. The problem comes when we try to extrapolate things like percentages to the population of units in the field. In the case at hand (BTW, the word you are searching for is "poll", which itself does not exactly apply in this case), the sample is biased to unknown extent, so that we can't place much confidence in statistics (like percentages) based on them. We have pretty high confidence that there are at least 24 defective units, and 15 working ones, but there is an awful lot of territory in between those two figures.
Data generally speak for themselves. The sample may be problematic, but unless the respondents are lying, the data (plural, by the way) contain some information. For instance, if 24 respondents report reliability problems, then we are pretty sure that there were at least 24 defective units in the field. The problem comes when we try to extrapolate things like percentages to the population of units in the field. In the case at hand (BTW, the word you are searching for is "poll", which itself does not exactly apply in this case), the sample is biased to unknown extent, so that we can't place much confidence in statistics (like percentages) based on them. We have pretty high confidence that there are at least 24 defective units, and 15 working ones, but there is an awful lot of territory in between those two figures.