I have seen several polls of this nature, and in my experience they have never worked.
Firstly, the questions are usually not tightly worded enough so that, as Alf suspects, different respondents interpret the question and grade their answers in different ways.
Secondly, since the raw data is what is presented, there is no adjustment for, nor recognition of, distortion in the data. There will be an over-representation of those with fewer purchasing experiences.** An amateur may place a higher importance on price compared to a professional who may place a higher importance on meeting a delivery date. And there will certainly be an over-representation of negative experiences.
Thirdly, the reader who does'nt appreciate the distortions in the data may gain a very false impression, whereas the reader who does appreciate the distortions will realise that the data doesn't actually tell you very much at all.
Finally, everything here is in the public domain, and I have seen more than one website running this kind of poll give up after attention from a company's solicitor.
** Consider person A makes one purchase, has a bad experience and awards 1. Person B makes 99 purchases, all are perfect, and awards 9. Company gets a rating of 5, 'Average', even though 99% of purchases were perfect.
Dave