Textmining – a modern method of market research
Textmining – a modern method of market research
10.02.2020
Qualitative techniques, such as in-depth interviews or focus groups, but also quantitative approaches, such as online or telephone surveys, are part of the standard repertoire of a market researcher.
This generates a huge amount of information, often only available in semi-structured or even unstructured form. How can those information be efficiently collected, analyzed and used?
As part of an online study, Facit Digital interviewed 1,000 German sports fans about the relevance of a "Smart Stadium". To get to know the target group and their needs as well as possible, respondents were asked to describe what information, assistance, entertainment programs or services they would like to have to make a stadium visit even more attractive to them?
You can certainly imagine how detailed the answer of an enthusiastic and demanding soccer fan may be.
First impression based on a tag cloud
To get a first feeling regarding the users’ comments, a wordcloud is a good choice.
First the raw data from the survey are cleaned up, i.e. numbers, punctuation marks and blank spaces are removed, uppercase letters are converted to lowercase letters. Stop words, i.e. words that have no additional value to the information content of the statement, are also excluded from the analysis. With help of the so-called tokenization, sentences are then broken down into keywords. As in the picture above, tokens can be words, but also expressions or whole sentences.
This is all done by using R, a free programming language for statistical calculations and graphics. R also offers the possibility to customize the created wordcloud, in other words the order, color and size of the words can be alternated with an appropriate syntax. Thus the wordcloud above represents a maximum of 200 words and displays only those that have been mentioned at least five times.
Getting to know the basic mood of the participants via sentiment analysis
This subsection of text mining describes the automatic evaluation of texts with the aim of recognizing an expressed attitude as positive or negative. This method is well suited to get a first impression of the mood picture regarding a specific design or form draft.
In another internal study by Facit Digital, the picture above (left) was presented to participants. The users were asked to comment on what they liked, disliked, or where they might have difficulties in understanding. With the help of a visual feedback tool (right ) it could be determined which areas were evaluated positively and which negatively. R also offers functions that allow an exact analysis of the mood after cleaning up the raw data.
With a value of "- 0.05" (rounded), the average sentiment score shows a slightly negative feeling. The market researcher then decides whether a more detailed look into the data is necessary in order to localize the specific strengths and weaknesses and derive well-founded recommendations.
