Definition: A matrix question allows respondents to select a single answer choice for each row inside of a question grid. Matrix questions are used when you have similar sub questions (rows) using the same answer choices (columns). The results for this question type will display response counts per each column and row with a resulting weighted average.
This question is commonly used when you want to rate a group of similar attributes. In the example blew, attributes of a company are displayed as rows, and the rating for each attribute is displayed as columns. This allows a customer to quickly give feedback for various topics using the same scale.
To create a matrix survey, simply create a survey as normal, and then add a matrix question where you see fit. An unlimited number of rows can be included along with up to twenty (20) columns. To limit survey fatigue, we recommend including no more than twenty (20) rows and ten (10) columns.
In the example above, we only include three (3) columns for the various levels of satisfaction. We recommend for any type of satisfaction rating, only a scale of three (3) or five (5) levels should be used. Anything more incuses survey length, increases survey fatigue leading to lower quality data.
One downside to a matrix question can arise if you’re are using it to determine the importunate of different attributes (rows). If a hotel chain used a matric question to how important (on a scale of 1-5, different attributes are (such as mattress comfort, parking, proximity to beach) nearly respondents will likely rate all of these as important. Ideally hotel management would want to identify what is truly most important to guest so they can have a narrow focus and use resources appropriately.
A solution to this problem is MaxDiff. MaxDiff is used to identify what is most and least important (or most/least desired) from a list of attributes. The basic concept is that respondents are shown a small subset of the total attributes (like a random set of five out ten attributes) and pick what is most and least important. Respondents are shown multiple sets, meaning attributes are compared against one another in.
Results for a matrix question will include the response counts per each column and row with a resulting bar chart. The bar chart can help you quickly identify trends. A weighted average for each row will also be computed for each attribute. You can change the value of weights at any time in the survey creation page. The Excel export will contain a column for each matrix row, with the rating value each respondent selected. If a respondent skipped a question row, that cell would be left blank.
Creating a segment report for a matrix question can help identify hidden issues. A regional store manager might want to segment results by state or city. The survey could atomically include store location using custom data. With this data a segment report (or a cross tabulation report) could be created. The segment report would then include the table shown above separated by state or city. The reginal manager could then use this analysis to follow up with certain stores to find ways they can improve certain areas.