How to decide the number of options to give to people in a poll? (for opinion based, subjective contexts) Just to be clear, this question is open ended on purpose, I am fine with(or even prefer to) being pointed to a few research papers, and proper areas related to this subject. I am not sure what I am asking is called, so I am very limited in my search.
Consider a general poll on a product to compare subjective choices:
"How important is the product design?"
"How important is ease of use?"
"How important is reliability?"
And a poll like this often has choice circles below them. For example, 5 circles are placed and picking 1 is "Not important at all" and 5 is "Very important"
An example in real life is vacation polls. Every hotel I stayed at had me fill such as poll at the end of my stay. Most of the questions being highly subjective, since their job is highly subjective as well. "good customer service" wasn't defined, just your satisfaction.
For such polls that are polling subjective matters/opinions, how is the number 5(most common I have seen) decided?
If it was only 3 choices, I imagine there would be a considerable amount of information lost. Is that the case? If so, to what extent?
On the other hand, I don't imagine having 100 choices would help them, since nobody can reliably say "I would rate the customer service in this hotel at 96% whereas the previous place I stayed at would be rated at 97%"
 A: When someone need to measure a variable that is not measurable using a well established instrument but is commonly agreed to have "low" and "high" values he/she might provide a set of questions (items) whose answers will describe the nature of the underlying distribution, and make him/her capable of comparing various groups of individuals. 
Concerning the interconnection of the items in such a questionnaire, there are 2 cases. Either the questions are independent from each other or they are organized in groups whose sum is represent a different face of an abstract notion (e.g. satisfaction). In the latter case the questionnaire is called a Likert scale, may consist of several "questions" - items (10 to 90 items or even more) and all of the items should use the same number and type of possible answers. 
In both cases the researcher or organization who constructs the scale should choose a set of possible answers to provide to the participants of the survey. The number of options varies from 2 to 11 (see some examples from 2 to 7 options here and from 4 to 7 options here). Two or three possible answers usually are not considered adequate since the participant may need more possible options to express properly his/her opinion although, for any practical (that is statistical) reason there is not significant difference when sums or mean values are calculated (see for example this article). Further, when the number of possible answers is increasing, the time of completion is also increasing thus it is seldom to see a scale offering more than 7 options (actually, after a short Google search I did not manage to find an example of a scale offering more than 7 possible answers).
It is true that usually five points are selected as a number that offering adequate answers without tiring the participants. However there is at least two disadvantages of this option since it is not always clear whether the middle point means "Neither agree nor disagree" or "Don't Know / Not applicable" (error introduced by misunderstanding), while it is also well reported that this option introduce "social desirability" bias (see for example this article). In contrary, 4 or 6 options was shown to be a better option regarding this "middle point bias" (see for example this article) having although a more aggressive form since they force the participant to take either a positive of negative position.
To summarize, there is not a right or wrong choice on number of options offered, there are pros and cons for every choice and usually there is also a tend to follow the selections made from former researchers in the field.
I hope that the above will helps. 
