Know-How Exchange : Research/Metrics

The second scale will provide you with more information (number of people who are undecided), as making the assumption that a number '3' on the first scale represents anything other than a midpoint of that scale is dead wrong.

Omar, the second scale is the 'better' one of the two, but the you should consider doing this:
Give option to select a number on the 1-5 scale OR to select 'undecided'. This way you can distinguish between those who are truly undecided and those who are decided but feel a '4' is too high and a '2' is too low.

If you are designing a research instrument I suggest you read this quick page avoid pitfalls of DIY market research.
https://www.revealresearch.com/learning_centre/articles/i/why_hire_a_profes...

good skill,
Luke

Dear Omar

It’s potentially more complicated than Luke has stated, depending on how well your respondents read the questions and how well they interpret the possible answers, which is usually not at all well, hence my proviso of “Potentially”

“Undecided” if you choose to use the term, means that they haven’t actually made up their minds. They could plonk for a 5 or a 1 or a 2 or a 3 or if they thought that 3 equalled neither satisfied nor unsatisfied, or neutral. It is in fact an answer akin to “don’t know” or “undecided” in a poll which provides a potential swing population should someone influence them further. They don’t have a strong opinion through experience and are therefore capable of moving in either direction. It is unusual to see this term in a survey as I can’t see the point in it unless you are seeing how large the sector which you can influence is, just by changing their perception of the product or service. Note: Not actually changing the offering. And that’s dependent on how accurately they read the questions and the choices of answers!!

If you choose the term “Neutral” instead, you set a mid-point for the survey, in which case everything which Luke has said falls into place. Not to define the set-point of neutrality is sloppy and dangerous as you cannot assume that respondents will make the same assumptions as you do.

You might find the differentiation between “Undecided” and “Neutral” or “Neither Satisfied or Unsatisfied” as an argument over semantics, but in surveys and polls the language of the question can so strongly skew the answers given that they can ruin the prediction sought. I can’t remember the actual question which they got wrong, but Mori, the pollster cocked up their forecast of the 1992 UK elections by a huge amount because of the phraseology of a single question. To their credit, they admitted their fault, rather than ascribe it to the vagaries of the poll!

Steve Alker

Also, consider adding a Not Applicable choice. If the question doesn't apply to my needs/background, having a way to opt-out of a question is a big plus.

Here is potentially another way although I think Steve has the best point about selecting a completely different way than a value scale.

Ask the question thusly

On a scale of 1 to ? (Customer picks the max value)

my experience with your company rates a ?
(customer picks the degree to which we successful met your expectations.)

It is easy to normalize all the data after the fact and lets people really think about everything that went into a value experience. You can even deal with negatives, yes some customers will give you negative scores.

Perhaps more simply put have your customer pick two numbers, one representing a perfect experience and the second representing how close we came to meeting all his or her needs.