Defintion: Maxdiff analysis is also called as the best-worst scaling. Maxdiff analysis is an analytic approach obtained to gauge survey respondents' preference score for different items. Maxdiff analysis is similar to conjoint analysis in many ways. However, the only visible difference being maxdiff is easier to use and is more comprehensive when you want to analyze critical research situations.
Message testing, customer satisfaction, brand preference, and product features are the common surveys where maxdiff analysis is used to predict the outcome. Maximum Difference Scaling or maxdiff question involves the ‘Best’ and ‘Worst’ scales in a given set. The respondents are asked to pick the most and least important factors in given answer options.
Maxdiff is used for
- Risk Management
- Security Policy
- Organizational Security
- Asset Management
- HR Security
- Physical and Environmental Security
- Cloud Computing
- Communications and Operations Management
- Access Control
- Incident Event and Communications Management
- Business Continuity and Disaster Recovery
Maxdiff example: A smartphone making company is interested to launch a new smartphone. However, before launching the new phone they are interested to understand what features, prospective customers are looking for and how the priority of these features impacts their buying decision.
The company conducts a maxdiff survey to understand the must-have features and to know the features that will make a difference. If a smartphone making company has a detailed knowledge of how customers perceive their preferences or attributes, they can create smartphones having a wide range of features that are of value to the customer.
Benefits of maxdiff analysis
- Powerful discrimination
- Easier to respond
- Test a larger number of items
- Eliminate scaling bias
- Allows diversity
- Get precise ratio data
- Implements tradeoff
- Brand preference
- Customer satisfaction
How to use maxdiff?
Before explaining how to use a maxdiff question, it is important for you to get acquainted with the common terms used in the maxdiff analysis question.
- Attribute: It is an individual statement that researchers like to rank.
- Set: It is a group of attributes. Researchers have the liberty to determine the size and number of attributes in a set that will be displayed to the respondents.
In a maxdiff survey, a set of attributes are shown to the respondents, and are asked to choose the ‘most’ and ‘least’ items in the attributes.
Maxdiff survey example
Here is a maxdiff example created using QuestionPro's online survey platform. In the adjoining steps you will understand how to use a maxdiff question in your online survey.
- Go to Surveys >> + New Survey
- Enter the survey name and press Create Survey button
- Press Add Question button
- Go to the Advanced tab and select the Max-Diff question type
- Enter the question and add the attributes by going to Edit Options in Bulk.
- Go to the Settings tab situated in the upper right-hand corner.
Here you can select the maximum attributes you want to be displayed. You can decide several attributes per task and also decide the attributes you want to be repeated. Additionally, you can also select whether you want to randomize the attribute display or not.
The number of task count shown to each respondent is based on the total number of attributes tested and the number of attributes you wish to display at a time.
- You can preview the question by pressing the Preview button present at the top right-hand corner.
- Test the maxdiff survey multiple times until everyone is satisfied. A well-tested survey gives you the assurance that the data you will be collecting is of value.
Maxdiff analysis example
After conducting the maxdiff survey, it's time to analyze the collected data.
Maxdiff analysis as a whole helps you understand why most of the respondents prefer a few attributes ‘most’ and a few attributes ‘least’.
- Go to Analytics >> Choice Modelling >> Maxdiff Analysis
- In the above maxdiff analysis example you will find a red square at the upper right-hand corner with three icons in it.
- Download the report in excel format by clicking on the XLS icon present at the upper right-hand corner.
- Download the report in SPSS Format by clicking on the SPSS icon placed beside the XLS icon.
- You can set filters to generate data specific to your needs by clicking on the Filter icon placed beside the SPSS icon.
Dashboard customization for maxdiff analysis
- Go to Analytics >> Dashboard >> scroll down
- Go to Chart Types icon placed at the upper right-hand corner
- Choose the type of chart in which you want to display the results
- You can display the maxdiff analysis in Pie, Column, Bar, and line charts
Maxdiff analysis Vs standard rating scale questions
Researchers have found out that rating questions are susceptible to user scale bias, scale meaning bias, and it lacks discrimination. In addition to that, ranking questions have limitations of their own like being over bias, difficult for respondents to evaluate, limitation in testing number of items, ordinal data obtained limits the analysis and it does not allow ties.
Constant sum questions also have the same limitations and loopholes. Furthermore, researchers have seen that when presented with constant sum questions, respondents try to make the task of evaluating all the items easier by engaging in the response strategies.
Having understood the problems and limitations of different survey question types; especially with rating scale questions, researchers prefer to use the Maximum Difference Scaling or maxdiff survey as we call it. We can call maxdiff analysis as a trade-off analysis technique allowing researchers to conduct pairwise multiple comparisons.
By effectively using the maxdiff question, researchers could ask the respondents to select the most and least preferred or important points from the given list of answer options they are interested to test for greatest difference amongst items.
- If we compare the maxdiff and standard rating scale questions, then the maxdiff question will show greater discrimination amongst items and between responses received for the items.
- In a typical maxdiff analysis example you will notice how easy it is to apply a robust trade-off technique compared to other rating scale questions.
- Compared to the rating scale questions, the resulting item score is easy to interpret in the maxdiff survey as it is placed on a common scale of ‘0’ to ‘100’ points and summed up to 100.
- Being simple to understand and easy to reply even for adults and kids, a maxdiff survey yields more reliable data.
- In a maxdiff survey, you directly ask respondents to make a choice and not to express their strength or preference using some numeric scale. Hence, there is no scope for scale use bias.
Look at the above image and you will understand the basic difference between a rating scale question and a maxdiff question.