Eight tips for using a word cloud in market research story finding
Ray Poynter
Word clouds can help you find stories when working with market research data. In this post, I use the example of an open-ended response to a survey question and highlight eight tips.
The context
The example I am using for this post is from a study I ran during the pandemic and the lockdowns. The open-ended question I am looking at was “If you were to learn one new thing over the next few months, what would it be?”. The data includes over 800 separate answers.
For example, I am using Word Cloud Plus, a tool which is free for basic word clouds, and which has an inexpensive premium version for more complex uses.
1 Clean the data before you start
In this data set, several things needed tidying. For example, some of the responses were not in English, there were spelling mistakes, and there were irrelevant responses such as N/A.
My go-to tool for cleaning data is Google Sheets. I paste the open-ended comments into Column A and sort the data alphabetically. By skimming down the data, once sorted, you will spot patterns that need attention, such as entries such as “N/A” or “****”.
To consolidate the text into one language, I use the Google function “GoogleTranslate”. In B1, if I want to translate everything into English, I type =GoogleTranslate(A1,”auto”,”en”). This detects the language in A1 and then translates the content of A1 into B1. The cell can then be filled down, translating all the open-ended comments in a single pass.
The final step is that I copy column B and ‘Paste Special’ as ‘Values Only’ into column C. I then select column C and use the Tools/Spelling, Spell Check option. The spell-checking process gets faster as you work through the file, provided you select Change All as often as it seems safe.
2 Create an initial Word Cloud to see whether you are on the right track
When using a word cloud to help with the story finding, we are not looking to create a presentation word cloud (although we might do that later). We are creating word clouds to help us find the story or stories in the data. This means it does not make sense to spend too much time on preparation; you are looking to move fast and then explore different possibilities.
Below is the initial word cloud from this example.
3 What impact has the question had on the Word Cloud?
When we ask an open-ended question, the words we use in the question will often be played back in the answers we receive. The words used in the question do not necessarily help us understand the message in the data. By looking at the Summary view of the word cloud, we see that ‘learn’ and ‘time’ are both frequently occurring words. For example, “Would like to learn more about power bi tools” and “For me it’s time to review the way I organize my work”. Similarly, we can see that the word ‘skills’ is simply a clarifier, as in “more data science skills”.
The word cloud can be made a bit sharper by designating ‘time’, ‘learn’, and ‘skills’ as stop words.
By contrast, the prominent emergence of Market Research in the middle of the cloud is a reminder of the domain that we are working in. It could be removed, but it is probably helpful to keep it there, to keep our thinking located in the context of market research.
4 Look at the underlying text to understand the messages
Whilst the word cloud provides a starting point for analysis. The word cloud is a simplification of the underlying data. When looking for the story, I move backwards and forwards from the Cloud to the Summary and to the underlying occurrences in the text. In the previous tip, I looked at the occurrences of ‘skill’, ‘learn’ and ‘time’ using the Explain link in the Summary view.
5 Check for alternative spellings
In this example, we can see that some people have typed ‘qual’ and some have typed ‘qualitative’. To get a clearer picture, it is a good idea to use Replace to standardise these two. Similarly, there are multiple occurrences of both ‘working’ and ‘work’, and the cloud is more straightforward if ‘working’ is replaced by ‘work’.
6 Try displaying different numbers of items
The default for Word Cloud Plus is to display 50 items. However, you will sometimes find that showing more or fewer items leads to additional insights into the messages in the data.
7 Use colours to group similar concepts together
In the example below, I have assigned stop words, used qual, quant and work as spellings, changed the number of items displayed to 40, and colour-coded some key groups.
The word cloud highlights several themes to explore, including learning qual, learning tech, and non-research topics such as the guitar or a language.
8 Use the Hermeneutic circle*
The Hermeneutic circle is a process of understanding text by cycling from the total context to the individual parts and back to the total context. When working with a Word Cloud to understand text, the cloud (the visualization of the text) is the total context. To gain an understanding, you need to keep visiting the underlying words to ensure that your interpretation is based on both the parts and the whole. In the case of Word Cloud Plus, this means cycling from the Cloud, to the Summary, to Text (via References & Explain.) You can read more about the Hermeneutic Circle by clicking here.
Want to try Word Cloud Plus?
All of the features I have used in the example above are part of the free version of Word Cloud Plus. To try it out, click here, create an account, and off you go.
*Hermeneutic circle, if you would like to learn more about the Hermeneutic circle, check out ‘The Art and Science of Interpreting market research evidence’ by DVL Smith & JH Fletcher.