I've explored the big research provider stance on Likert scale questions and along with my views compiled 12 things to know about them before using them in a survey. It was developed in 1932 by the social psychologist Rensis Likert - but I expect you're not here to find out that!
Question selection - Likert scales are used to gain measurable data from specific statements. These questions are respondent-friendly and limits bias if designed correctly.
Question design - Likert style questions produce ordinal variables which are ranked categorical measures. This means the distance between responses is not numerically meaningful, strictly speaking, even when there are numbers assigned to responses.
Question design - Radio button grids, a drop-down menu, or a sliding scale are question types you can use for Likerts.
Question design - A Likert scale is a question which usually uses an odd number of options on a scale - typically a five-point or seven-point scale, with the middle option as a 'neutral' option. However some survey designers want to 'force' an answer so use even scales but this can put off respondents if they feel they cannot answer truthfully.
Question design - I like to include 'not applicable' to those who don't feel they have a stance on a topic (hopefully this is far and few between!). I don't feel 'neutral' and 'not applicable' are the same thing.
Question design - Outside of typical agreement scales, it’s better to use a unipolar scale that ranges from “extremely happy” to “not at all happy”, rather than a scale that ranges from “extremely happy” to “extremely sad”. Unipolar scales are just easier for people to think about, and you can be sure that one end is the exact opposite of the other, which according to SurveyMonkey makes it methodologically more sound as well.
Question design - Make sure scales are unambiguous - for example 'Never—Seldom—Sometimes—Often—Always'. How do you quantify the difference between sometimes and often? Is seldom universally understood?
Survey design - Use positive and negative indicators for balance. Some say that there should be an even number of both types of statements. But if you are doing a negative and positive for every statement then your survey can get long... most surveys I see have a mix leaning towards positive sentiment statements than negative. A mix is deemed vital though to limit bias and identify speedy responders who select conflicting answer options.
Analysis - Averages, medians, and frequencies are the tools you need for analysis. The tendencies in the data will give you answers to better answer your research objectives. You can group answer options to better understand negative and positive sentiments such as 'Net agree' or 'Net disagree' buckets.
Analysis - Answers in Likert scales can be used to understand other numeric based questions via key driver analysis - assign a number to each item on the scale and use those numbers to attribute those to Net Promoter Score, Customer Satisfaction or customer effort. This can be achieved via key driver analysis. There are also instances where means, standard deviations, t-tests, ANOVA, and other parametric statistics can be used on Likert style questions and scales. Strictly speaking, interval measures have numerically meaningful differences between values (e.g., number of visits, scores on the maths test). However, in practice, it can sometimes be useful to include means and standard deviations for Likert questions.
Analysis - If a lot of your responses are indifferent do not forget this group - what makes say 'neither' do you think?
Visualisation - Once you have your results, Likert scales can be designed visually by a stacked bar chart such as that below. The chart below is specifically highlighting those that have agreed with the scale 'I am paid appropriately for my role'. Traditional bar charts can be used but take up more space on a page.
Alternatively, select a statistic from the data to present to help tell a story...