# Yes, There is a Right and Wrong Way to Number Rating Scales

→
Yes, There is a Right and Wrong Way to Number Rating Scales

Once we’ve identified a specific construct of interest, the question becomes; ‘how many scale points?’ There’s an easy way to decide.

All constructs are either bipolar or unipolar and the number of scale points we assign to a question depends on whether its hidden construct is bipolar or unipolar. Bipolar constructs are those where attitudes can fall on one side or the other of a midpoint that itself is true ambivalence or neutrality. For example, a school principal might want to know if parents think the school day is too long, too short, or just fine. Length, then, is a bipolar construct.

So far that’s three response options that we can easily wrap our minds around. After that a response option in between the midpoint and the anchors is about all our brains can handle. Thus we are left with: Extremely helpful, Very helpful, Somewhat helpful, Slightly helpful, Not at all helpful–five scale points. While there is a “middle” point, the expression of that point is decidedly not synonymous with “neutral.”

Bipolar scales require seven scale points, three around each side of the midpoint–again a midpoint that truly means neutral, neither, or both.  If we go back to our school principal example, she would ask if the school day was: Much too long, Somewhat too long, A little too long, About right, A little too short, Somewhat too short, Much too short.

There has long been a ton of debate and confusion about scale point numbering. The academic scholarship has demonstrated that scales are most reliable when constructed with five and seven scale points.

Have questions on scale points or if the construct you are trying to measure is unipolar or bipolar? Ask away in the comments below.

#### Do a Better Job with SurveyMonkey Enterprise

Give yourself—and the rest of your team—the answers you need to make smarter decisions. Access our most powerful tools with SurveyMonkey Enterprise.

• Dan Hogan

Should your choices be listed from negative to positive or just the opposite?

• Phil G

Don and Paul–Indeed there is the possibility of primacy effects in questionnaires presented visually whereby people tend to favor the earlier response options. Stay tuned for our posts on avoiding order effects using flipping and rotation.

• Paul Therrien

Hi Dan,
There probably is no right or wrong answer. I would think that since our North American culture usually proceeds from left to right when reading, graphs are usually “0” on the left moving to more on the right, I would build my choices from negative to positive if it is an horizontal configuration and from positive to negative if the the configuration is vertical.

• What is your opinion of the effect of not offering a mid point, say with 4 scale points? This potentially forces answers which demand either positive or negative atitudes

• Phil G

Ian–Thanks for your question. There is no “midpoint” per se in a five point scale because its not “one side or the other,” but rather an amount. Think of the construct ‘likelihood’ where the middle option is actually something like a 50% likelihood—which in the case of rain, might cause you to bring an umbrella to work. In any case, it’s not ambivalence. As far as bipolar scales where there is a midpoint, you want to let people give an ambivalent answer if that’s what they mean, anything else introduces noise into the parts of the scale where true answers are.

• Hi – Are you saying that a bipolar scale should be 7 points and not 5? How about:
Very satisfied, somewhat satisfied, neither satisfied nor unsatisfied, somewhat unsatisfied, very unsatisfied.
Is that bipolar? Suggestions for improving it? What about saying “neutral” instead of the middle mouthful?
Thanks

• Michelle

2 questions – we typically name rating scales rather than asking for a numerical rating as we believe this helps the respondent more accurately answer. But some clients also want an overall rating ‘out of 10’ so we then apply a weighting to all responses. Do you have any recommendations as to how to apply a 0-10 rating scale to a 5 point scale?

further, there are clients that we know will more likely be rated highly on satisfaction scales. To increase granularity to results, rather than using a 7 point scale which we find too large and onerous, is it appropriate to have an ‘unbalanced’ scale such as dissatified, somewhat satisfied, satisfied, very satisfied, extremely satisfied. If so, then again we have difficulty in assigning a 0-10 rating later.

• Abid O

Looking forward to responses on the 11-point Net Promoter recommend scale, of 0-10.

• Actually science shows 10 point scales are to be preferred from a purely statistical view, because the data becomes less scewed.

However how is it to just use a 5 point scale from “Complete disagree —> completely agree” and then leave a 6 = Doesn’t know,

That way u can treat 6 as “missing value” in spss, or whatever. And 1-5 = a score.

?

• Teresa

I want to measure frequency and importance of each item. Can I have two scales for the same item; each would be 5 points from 0-4.

• Hi Teresa–If you want to measure frequency and importance, creating two separate questions would be best. Each question would then have it’s own rating scale, which should be fully labeled (e.g. Extremely Important, Very Important, Moderately Important, Slightly Important, Not at all Important). Let us know if you have any questions or if we can help in any other way!

• Patrick

What about rating scales? To compare different on-line lectures, if I ask for an overall rating, is it better to have a 0-10 scale, a 1-10 scale, or a 0-5 scale? I like 0-10 because people can get their head around base-10 numbers, like percentages.

Thanks!
Patrick

After two weeks of the service provided from 1-10 with ten being the best,
1. How likely is it that youb would rate “ABC Company” service or product to a friend or relative.
2. Sort into three groups, 1)9-10=promotors 2)7-8=passive 3) 0-6 Detractors
3.%of pormotors – % of Detractors = Net Promotor Score (N.P.S.)

What are your thoughts on this process?

• JAL16

Do people use questions that allow the surveyor to provide additional free form comments or are the questions typically just the format that I’ve seen thus far in the examples? Like if there is a response for “other” then they can explain in detail? If so, how do you evaluate the responses?

• Michelle

The main reason for my comment is to say I enjoy reading peoples comments. Comments to me are somewhat a helpful joyful therapy. I myself either like something, laugh to myself, walk away, or find a way to make it work for me.

• I’m constructing a survey for an organization around communication, culture, etc. It seems like the questions should be bipolar, but your question bank shows all these as unipolar. For example, it seems like I should ask: How effective is XX at communicating across the organization: Very Innefective, Somewhat Innefective, Neither Effective nor Innefective, Somwhat Efective, Very Effective. With a unipolar scale, there is only one negative response and four somewhat positive responses. Won’t that cause bias, and not be helpful in uncovering poor communication?

• Phil G

Hi Gerry,

Thanks for your question. The basic rule is that for most constructs, in-, dis-, un-, etc. do not automatically create cognitive symmetry. What is cognitive symmetry in simple terms? That is, the prefix does not immediately create an equally balanced antonym for the word of interest.

For example, let’s use the construct of interest. It isn’t possible to have less than ‘no’ or ‘zero’ interest. Therefore, uninterested isn’t the opposite of extremely interested, very interested, somewhat interested, or slightly interested.

Hope this helps.

Phil

• Hi there! I’m at work browsing your blog from my new iphone 4! Just wanted to say I love reading through your blog and look forward to all your posts! Carry on the superb work!

• Thanks so much!

• Gavin

I am interested to what extent teachers use culture learning strategies. I am considering using a 5 point scale from “not at all”, “not often”, “sometimes”,”often” to “to a great extent”. Is this the right way to go in your opinion? I have also considered “never” and “always” as the extreme choices.

• Thanh

Hi Palle,
Disagree and agree survey would be bipolar, wouldn’t it? Since it is not a knowledge question, it is not wrong or right. If the person doesn’t know, it means neutral, still in the survey, exactly a 5 in 0-10 scale or 2.5 in a 1-5 scale.

Hi Patrick,
Just an idea. How about you have 6 questions, 1. Rate the audio quality (you can split this to 3, speaks too fast? too low? weird accent?), 2.Rate the picture quality, 3. Rate the content usefulness, 4. Rate the content attraction, 5. Rate the loading, 6. Put in order of importance to you, audio, picture, usefulness, attraction, loading.

Hi Gavin
“Not often” sounds the same as “sometimes” to me. Is “rarely” better?

• Thanh

Hi Patrick,
Complimentary to what I said. I still don’t know how my “method” can calculate the overall score, sorry. Why do you need it? You can look at each aspect to make improvement.

• Stephen Okeyo

Great reading. Fortifies my understanding of how to make quantitative data from qualitative information

• Amazing layout I think.

• Wonderful work! That is the kind of info that should be shared across the web. Disgrace on the seek engines for not positioning this submit upper! Come on over and talk over with my web site . Thanks =)

• Hanna J

Thanks for your support! We appreciate you 🙂

• Fran

Is there a way to prompt for a comment by the participant if they score something low? Example: On a 4 point scale, they mark something “good” with excellent being the highest sciore. Could it prompt with a comment, such as… You gave the instructor a score of “good”. Can you tell us what he could do to improve his score to “excellent” next time?

• Benr

Hi, I have been reading some research that uses a four point rating between 0-1 with the followiing intervals:
lowest – 0
second – 0.333333333
third – 0.666666667
Highest – 1

I intend to use the same for my question ratings – do you think this is ok?

Thanks

• Hanna J

Hi Ben – We recommend that you use words when possible to label your rating scales. See this blog post for more information: https://www.surveymonkey.com/blog/2012/01/24/words-speak-louder-than-numbers/

If you must use numbers, and need them to be on a 0 through 1 scale, can you use a 5-point scale and make them percentages? 0%, 25%, 50%, 75%, and 100% seem like much more intuitive responses to us than 0, .333333333, .666666667, and 1. What do you think?

I would argue that “Somewhat too long” and “A little too long” are equivalent responses in the minds of most users, thus the 7-point example given above could be reduced to 5 points. Adding responses to your scale that have no meaningful difference between them just dilutes your data.

• Hanna J

Hi Adam – You’re right, you want to try and get the information you need through questions that are as succinctly as possible. Our research shows that many people differentiate significantly between “somewhat too long” and “a little too long,” but you always want to use your best judgment. Thanks for sharing your insight!

• Sylvie Potoniec

I appreciate all the great information. I am using a 7 point scale with a Not Applicable choice. Will the results be accurate?

• Hanna J

Hi Sylvie – We recommend using a 5-point scale if your what you’re measuring is unipolar and a 7-point scale if what you’re measuring is bipolar. It sounds like yours is a 6-point scale with a n/a option which we do not recommend. We use odd numbered (5 or 7) scales because it allows for a midpoint. We also do not recommend supplying a not applicable choice unless you are specifically interested in those who feel the question is not applicable as it may lead to people simply avoiding filling out your measure. A better way is to use skip logic so that those for whom the question does not apply never see it in the first place. Does that help? Let us know if you have any follow up questions. Thanks!

• Sylvie Potoniec

Just to clarify, I am using a 7 point scale and measuring is bipolar. I will look into skip logic!

Thank you!

• jackie a

how do you identify the 1-10 scale with out qualifing the high and low ie…
“How would you rate the meal on a scale of 1-10, 10 BEING THE BEST?”
In this example the phrase “10 BEING THE BEST” should be able to be replaced by one word that describes your scale. Alternatively there should be a different word that describes a scale where #1 is the best. The end question would be phrased,
“How would you rate your meal on a ______ scale of 1-10?

• Kayte K

Hi Jackie, we recommend keeping your scales to 7 points instead of 10. Hope that helps!

• Isha

I am not clear about polar scales (negative/positives) and polar scales (non neg/pos). Please suggest.

• Kayte K

Hi Isha- we’d be happy to help! If you could kindly send a note to support@surveymonkey.com, they can help you get started in the right direction.

• Brandon B

When thinking about the “length” of an event, “much too long” and “much too short” are really both negative, so why would a bipolar scale work? For instance your values are -3, -2, -1, 0, +1, +2, +3. How do you decide which is -3 and which is +3? In my opinion this scale does not work for “length” but usually works for all other measurements such as “agreement” or “expectations”

• Steve

Can anyone point me to the academic scholarship that demonstrates the improved effectiveness of 5 and 7-point scales? I’m being asked to change our existing 5-point scale to a 10-point scale, and if I’m going to push back at all, I need some hard facts.

• ellen konar

Krosnick and Fabrigar (1997) “did not find a monotonic increase in reliability as the number of scale points increased. Instead, a curvilinear pattern emerged in their review such that scales between 5–7 points were more reliable than scales with fewer points or more points. This was true for both bipolar and unipolar scales.” Typically, the rule of thumb is 5 pts for unipolar (e.g. never … always) and 7 pts for bipolar (e.g. very dissatisfied to very satisfied)

• Alexander Suraphel

Ellen, I guess I’m too late in the discussion but I’m considering a three point scale for something else to make ppl rate more. How do I assess it’s effectiveness?

• MFsurveymonkey

Hi Alexander! I’m not Ellen but I hope I can help 🙂
A three point scale will lead to less variation than 5 or 7 point scale simple because respondents are provided with fewer choices.
For example, a 5-point scale will show strongly disagree, somewhat disagree, neither, somewhat agree, strongly agree. So if you reduce it to a 3-point scale, that means people who would originally choose somewhat agree will be forced to choose strongly agree or neither. That way, it’s likely that you’ll see more agreeable answers.

• Alexander Suraphel

MFsurveymonkey, thanks for your reply! I’ve thought about that too. My problem is that I needed to make a tradeoff. Most people postpone doing a rating for more than three point scale. Anyway I’ve chosen to use the 5 point scale for the reasons you mentioned.

• MFsurveymonkey

That seems like a wise choice 🙂
Let us know if you have any further questions. Happy surveying!

• KTsurveymonkey

So glad it was useful to you! Happy surveying.

• Nick

Hello every one, I am a Greek postgraduate student, so forgive me if my English seem poor.

My question is this: How can scales with sample items like “I am fully satisfied with my job” (job satisfacton)

or “I am dissapointed that I ever entered the teaching profession”

or “I am capable of engagong even the most difficult students in my classroom activities” (self-efficacy)

or “My principal behaves in a manner thoughtful of my personal needs” (Leadership) can be best measured? Usually they are treated as bipolar constructs (from strongly disagree to strongly agree) but I have seen them as unipolar also (measured by the level of agreement with the statements). In your opinion, wihch of thes two should I use? My first choice was to measure with “strongly disagree to strongly agree” in a 5 point scale but now that I have read this article I am not so sure. I also plan to do regression analysis with my data so which choice do you think will provide the most accurate results?

• KTsurveymonkey

Hi there, Nick!

That’s a really good question. In this case, I’d say to use the scales that are most common in your area of research, so that you can have comparable data. So if previous researchers in your field have more commonly used 5-point bipolar scales, I’d stick with that. If unipolar scales that use the main verb in the question are more common, then it’s better to have a unipolar scale. That way, you can compare his results to those of other people in his field.

Agree/Disagree questions can be tricky, but if your industry standard is agree/disagree, those always should be 5-point bipolar (ranging from strongly agree to strongly disagree as he mentions below) and not 5-point unipolar. Hope that helps out, Nick!

• Nick

Thank you very much!

• Leah

Hi there… I have already altered my survey to be a 10 scale instead of 5. But now I’m trying to figure out how I convert the earlier responses received to that 10 scale. Tips please?

• KTsurveymonkey

Hi there, Leah!

We may need to look at your survey a bit closer to provide you some feedback on this. Would you mind contacting our support team here: http://help.surveymonkey.com/contact-us. Thanks!

• Leah

Thank you anyway but we’ve decided to stay with a 5 scale rather than the 10.

• Angela

I’m trying to decide between a scale of 1-5 of “Strongly agree, agree, neutral, disagree, strongly disagree” or 1-7 “Strongly agree, agree, somewhat agree, neutral, somewhat disagree, disagree, strongly disagree”. What would be better for getting honest opinions without confusing people? What is better for analysis afterwards?

• MFsurveymonkey

Hi Angela,

With that said if your survey is requiring a higher degree of precision, you might want to go for a 7 points scale. It’s up to you!

• Patrick

The relativeness of just “agree” or “disagree” could be an issue. What is often done is to label 1 and 7 as “completely…”, and 2 and 6 as “mostly..”

• Patrick

If you are recommending 5 or 7 point Likert scales with labels for each point, then why is the Net Promoter score question a 10 point scale with no labels?

• MFsurveymonkey

Hi Patrick! This is such a great question and a very valid point! IT’s important to realize that the NPS question type is on a league of its own. It’s done on a 10 point scale because it allows for more range, since the results are intended to fall between 3 categories (Detractors, Passives and Promoters), which already act as the labels. The 10 points allow for more flexibility in scoring, and produces more accurate results in this formula. It works very well for this specific topic (brand promotion), but if you’re looking to use a simple Rating scale for other topics then a 5 point or 7 point scale will be the way to go, as described in this article.
If you’re interested in reading more about the NPS, I can recommend some more articles, like: https://www.surveymonkey.com/mp/net-promoter-score/

• Patrick

So NPS accepts the risk people will answer randomly in exchange for greater granularity. I guess best not to have many 10 point scales in a single questionnaire. If I did want to have one, is there a way to set it up as boxes like NPS is layed out?

• MFsurveymonkey

Hi Patrick! Unless it’s the NPS question, then it might indeed be better to use the 5 or 7 point scale in your survey.
https://www.surveymonkey.com/mp/net-promoter-score-calculation/
I hope this is useful!

• Susana

NPS is actually an 11 point scale, using 0-10 to rank likelihood to recommend.

• X3Charlie

I’m researching this because I was looking into a voting system called range voting where people rate the candidates. Personally I dislike any scales above ten since I rarely vote below 6. Right now I’m thinking of a 3 point system which is actually 4 because you can leave it blank next to candidates you don’t like.

• MFsurveymonkey

Hi Mateja! This is a super popular question! Since you don’t need to get super specific, I think you should go with your instinct and use the 5 point scale. Our Likert scale would be mentioning 5 points as well. You can read more about this just here: https://www.surveymonkey.com/mp/likert-scale/