Quantitative versus qualitative data

 

In an over-simplified description we might define the following:

  • Quantitative data – tables of numbers such as can be found in an Excel spreadsheet
  • Qualitative data – text data as can be found in the written transcripts of interviews or focus groups

 

In reality the line between these two types of data becomes exceedingly blurred as new types of data are identified (eg. new types such as photos, videos, computer apps, etc) and new methods for analysing such data are developed (eg. can a computer extract the meaning from a written interview just as well as a person can?)

 

In the choice between quantitative and qualitative data I would often encourage using both types. The advantage of pure quantitative data is that it is collected in a purely systematic manner (people are asked the same questions using the same wording and the same ways of responding to a survey question), while the disadvantage is also that data is collected in a purely systematic manner (what happens if I want to provide a response with more detail than is allowed in the designed response types). Indeed within the Analytics space we often focus (sometimes exclusively) on quantitative data but I would argue that qualitative data is just as important (qualitative data will contain information on what types of questions should we consider addressing through Analytics, and on how different people interpret Analytics results – in turn potentially suggesting new features that could be added to Analytics results, new datasets to be incorporated, or different ways of reporting Analytics results).

 

Data visualization is well defined for quantitative data, but not so much for qualitative data. As such there are two approaches that I use for visualizing qualitative data:

  • Visualization types including word clouds (https://en.wikipedia.org/wiki/Tag_cloud) and concept maps (https://en.wikipedia.org/wiki/Concept_map)
  • Alternatively I wonder whether every time we collect a sufficient amount of qualitative data that we will always convert that qualitative data into quantitative data (eg. we might hold 100 interview with employees about job satisfaction, while in the initial interviews employees might be asked to describe their level of satisfaction in words we might then classify each interview as “mostly positive” or “mostly negative” through our manual efforts, and then analyse and report the data as quantitative data). In such a conversion process I need to understand what types of features am I looking for in the qualitative data, and what rules will I use for converting qualitative data into response categories (eg. positive or negative). Once I understand the rules that I am using for this conversion then I can consider extending these efforts to automate this process.