Jisc banner

What should I be aware of when using grid and scale/rank questions in my survey?

Both the grid question and scale/rank question types use a tabular (or matrix) format to display question statements and answer options.

While grids and scale/rank questions can be flexible and space-saving, care should be taken when using these question types in your survey, for the following reasons:

Multiple or large tables can affect how quickly a survey page displays to respondents

A moderately-sized scale/rank question containing 20 statements and 7 response options contains 20 x 7 = 140 individual elements.  Your respondent’s browser must obtain this information from the Online surveys servers, interpret it and then display it on a page.  How quickly this happens depends on the capability of your respondent’s browser and the available memory on their computer or mobile device.

The more grid or scale/rank questions you include on a single survey page, the harder your respondent’s computer has to work to display all of the information.

The larger the question, the longer it will take for it to display on your respondent’s screen

It is not just the number of rows and columns in a grid that determines how large your question is.  Including selection lists in your grid will increase the size of the grid object significantly.

For example, a grid question with 30 rows and 6 selection list question columns, where each selection list contains 60 answer options, comprises 30 x 6 x 60 = 10,800 individual elements.  A grid of this size will cause a survey page to load slowly.  It is possible that some respondents’ browsers will not be able to cope with such a large object and will fail to load the survey page at all.

Submission at the end of the page that contains a grid of this size will also be slow as a large amount of information must be transmitted from the respondent’s browser to the online surveys servers before the respondent can continue.

Multiple or large tables can affect how long it takes to edit your survey

When you edit a grid question or a scale/rank question, online surveys checks all rows and columns to ensure that the update is applied correctly and all rows and columns remain intact following the change.  The larger the question, the longer it will take online surveys to cross-check all elements in the table.

Multiple or large tables can be challenging to analyse

Each row/column combination is displayed separately in the analysis.  A grid with 10 rows and 3 question columns will produce 30 separate analysis tables or charts.  Depending on the question types included in the grids, and whether a coded or uncoded export is chosen, the exported files for surveys containing large or multiple grid questions can contain many hundreds of columns of results data.

Multiple or large tables are not accessible to respondents with disabilities

Questions presented in a tabular format can cause challenges for participants with cognitive disabilities and for participants using screen readers.

Multiple or large tables are cognitively demanding for respondents

There is some evidence that grid or matrix questions produce less than optimal results, and including multiple or large tables may even cause respondents to drop out of the survey.

What are the alternatives?

  • Split large grids or scale/rank questions into several smaller grids or scale/rank questions.
  • Split survey pages so that there are fewer grids or scale/rank questions on a single page.
  • Split large surveys into several smaller surveys that can be chained together using piping.
  • Consider whether all of the columns in a grid are essential – e.g. do you need a free text question column or can you ask for additional information in a single separate free text question.
  • Consider using validated free text questions instead of selection lists – e.g. a free text question limited to whole numbers between 1 and 60 rather than a 60-item selection list.
  • Consider using different, non-tabular, question types – e.g. several individual multiple choice questions rather than a scale/rank question.