Quantitative research is focused on numerical, quantifiable data and using scales is an important tool to collect nuanced insights. Designing an effective scale requires a researcher to think about the objectives, the best type of scale, and how they want to quantify the labels. To design an effective and engaging scale, listed below are a few do’s and don'ts, as well as, tips on designing scales.

Types of Scales

Scales are a way to collect nuanced measurements in research. Commonly used scales include: Likert, NPS and Semantic Differential (bipolar). While each has its own purpose at the core they are very similar. You assign a numerical value to each point on the scale which allows you to quantify feelings and opinions.

NPS

This type of scale is most commonly used in customer satisfaction surveys. Expressed as a numeric 10-point scale it allows you to get a quick read on your customers at any stage in their customer lifecycle.

Likert Scale

This scale is typically expressed by levels of agreement, but the possibilities for labels are endless. Respondents can select across a 5 or 7-point scale and typically include a neutral or moderate mid-point. Although this is a hotly contested topic which I will talk about in a minute.

Semantic Differential

Commonly referred to as a bipolar scale it’s designed to measure the connotative meaning of objects, events, and concepts. It differs from the other two scales as labels are expressed as opposite words or phrases, night owl vs early bird or slow vs fast. Each scale can serve a different purpose within a survey but all will help collect deep insights and help quantify a respondent’s feelings.

Design an Effective Scale

Before you design a scale, you will want to ask yourself the following questions:

1. How many points should my rating scale include?
2. Do I want to offer a moderate or neutral response option?
3. How should I label the response options?

Likert Scales are flexible as they can be expressed as 5 or 7 points with any set of labels. When deciding how many points consider your construct – is it bipolar (i.e. satisfaction) or unipolar (i.e. effectiveness). While levels of satisfaction are more nuanced and require more scale points it doesn’t make sense so measure negative effectiveness, so a 5-point scale would suffice.

Another thing to consider is how many respondents will complete a survey. If you are only surveying 200 people a 7-point scale could be excessive and you will likely combine top two and bottom two boxes to have enough sample to make a decision, so a 5-point scale would be sufficient. If you are planning to collect 1,000 respondents, you will have plenty of data to analyze all 7 points along your scale individually. In either scenario you could include a 7 point scale but to have enough sample you will likely combine top two and bottom two boxes.

Offering a neutral or moderate option is a hotly debated topic in research. Some complain that offering a neutral option allows respondents to be lazy and not make a decision and are disguised as a “don’t know” answer. However, in some cases you do need to provide a neutral option because forcing a respondent to choose could skew the data and introduce bias. Remember you are in control of the “points” so you can choose to either assign neutral a 0 or incorporate into the scale. When labeling answer points, it’s important to make sure the question text matches the label text. Don’t ask them if they are ‘likely’ to go to Target this weekend and then express the answers as agree and disagree. Instead it should be expressed as not very likely to very likely. This creates a complete thought for the respondent, “I’m very likely to go to Target this weekend!”.

Do’s and Don’ts of Scale Design

When designing a scale question follow these truths…

1. It should be easy for the respondent to interpret the meaning of each scale point.
2. All respondents should identically interpret the meaning of labels.
3. The scale should give enough points to differentiate respondents as much as validly possible.

Avoid these dangers:

1. If your first scale is negative to positive stay with this construct throughout the entire survey.
2. Don’t skew the data, each scale should have a set of balanced answer options.
3. Don’t overload your survey with scales. Respondents don’t want to answer 50 similar statements against the same Likert scale, you will lose their attention

Expand Your Horizons

While each scale looks different and is used for a different objective you can boil them down by thinking about them simply as a 3,5,7 or 10-point scale. So, the next time you want to incorporate an NPS why not use a Star Rating [link] question? It will still provide you with that 10-point scale data but the user experience is way more fun. This is especially helpful if you have multiple scale questions in a survey, it will break up the monotony. If you need to use a Likert scale type consider giving the labels an upgrade (i.e. Absolutely! to Absolutely not!), think beyond agree vs disagree and what your target market will relate to or even better relate the scale to your brand. Even these two minor tweaks will have a significant impact on the user experience and take your survey from good to great!