Most survey advice is about the stem, the sentence with the question mark. Fair enough, wording matters. But a respondent does not answer your sentence in a vacuum. They answer it through the options you give them, and those options quietly decide what they are able to tell you. A perfect question with a careless scale still produces careless data. The scale is not packaging. It is half the measurement.
How many points, and why it is not just taste
Too few points and you flatten real differences, forcing people who feel quite differently into the same bucket. Too many and you ask for a precision nobody holds, the difference between a 6 and a 7 on a hundred-point feeling. For most attitude items, five to seven points capture the range people can actually distinguish. The right number is the one that matches how finely your respondents genuinely experience the thing you are asking about, not the one that looks most impressive in a table.
Label the points, do not just number them
A scale that labels only the ends and leaves the middle as bare numbers asks every respondent to invent their own meaning for a 3. One person's 3 is mild agreement; another's is a shrug. When you can, label each point in words, and keep the spacing between those labels even, so that strongly, somewhat, and slightly mean roughly equal steps. Verbal anchors do not just look tidy. They make two people's 4 mean closer to the same thing, which is the whole point of a scale.
The midpoint and the missing option
Two small choices carry a lot of weight. A midpoint gives honest fence-sitters a home, but it also lets people duck a question they would rather not answer, so include it when neutral is a real position and think twice when it is an escape hatch. And when some respondents genuinely have no basis to answer, a missing not-applicable or do-not-know option keeps them from guessing, because a forced guess is noise, not data. Leave it out only when you are sure everyone can answer.
Where the Survey Development System keeps the scale honest
The reason scales get treated as an afterthought is that most tools let you attach any options to any question without comment. ReliCheck's Survey Development System looks at the fit between what a question asks and how it lets people answer, and flags the mismatches, a scale that cannot capture what the stem is really asking, before you launch. That check matters because reliability and validity both live partly in the response format, not just the wording, and a mismatch there is a limitation you would otherwise discover only in the data. ReliCheck is built to catch it while it is still a quick fix, so the answer you get back is the one your respondent actually meant to give.
ReliCheck's Survey Development System checks whether a question's response format fits what the item is asking, and flags mismatches before launch. See it at relichecksurvey.com.