@SachaG, I am thinking networks would be a good way to visualize results after feature and reason have both been normalized. See this link for a sample which uses a subset of the responses.
One possible improvement is to set different colors for feature and reason nodes (nodes are the circles). Circles here correspond to frequency or the number of people which indicated the same response. Edges are thicker if more people mentioned that reason in relation to the feature. Arrow origin is the feature and it points towards reason/s.
If you look at the link, you can see that select is a common feature respondents are unable to use (large circle or node) and the most common reason stated is styling (thickest edge coming from select). It can also be observed that styling is a common problem since a lot of other nodes point into it! Many other insights can be derived!
Network is interactive and you can hover to highlight relationships you wish to see. I used python here but library uses d3 under the hood.
Pain point I am seeing however is parsing because a lot of respondents did not follow the expected format of answer. In this vis for example, i was only able to process 600+ out of 1700+ responses because those were the only ones which are easy to parse (followed expected format of response).
@SachaG, I am thinking networks would be a good way to visualize results after feature and reason have both been normalized. See this link for a sample which uses a subset of the responses.
One possible improvement is to set different colors for feature and reason nodes (nodes are the circles). Circles here correspond to frequency or the number of people which indicated the same response. Edges are thicker if more people mentioned that reason in relation to the feature. Arrow origin is the feature and it points towards reason/s.
If you look at the link, you can see that
selectis a common feature respondents are unable to use (large circle or node) and the most common reason stated isstyling(thickest edge coming fromselect). It can also be observed thatstylingis a common problem since a lot of other nodes point into it! Many other insights can be derived!Network is interactive and you can hover to highlight relationships you wish to see. I used python here but library uses d3 under the hood.
Pain point I am seeing however is parsing because a lot of respondents did not follow the expected format of answer. In this vis for example, i was only able to process 600+ out of 1700+ responses because those were the only ones which are easy to parse (followed expected format of response).
Let me know what you think.