Not a data scientist, but forum polls are rarely useful data. They’re easily socially engineered and influenced by a small number of adamant voices
We have a User Research group whose job it is to perform blind and scalable measuring of sentiment and feedback, and present that information with as much personal bias stripped out as possible.
Forum, reddit, social media feedback is most useful as anecdotal evidence, and in some cases provides context in the form of volume of conversation. That’s not to say it isn’t useful; it’s simply more qualitative than quantitative