Background
Applying for product management internships is usually the same series of events. Link a resume, maybe a few fill in the blanks on a survey form, and submit. However, one of the most interesting PM applications I came across was IGN’s. Applicants were only given a set of questions to answer; no resume accepted. The process was great practice; this question was as follows…
“IGN has been collecting feedback from our wiki users to figure out ways we could improve their experience… Create a pivot table grouping this feedback into categories that will help us improve user experience on wiki pages, and what you would suggest for next steps.”
_Disclaimer: I assume this isn’t a breach of privacy, these questions are available on their application’s website _IGN Code Foo 2020
Overview
I decided to answer this question about grouping feedback from two angles (the lockdown gave me _a lot _of time to think); the straightforward “logical” approach using Excel and pivot tables, and the moonshot “creative” approach using machine learning & TensorFlow attempting to detect sentiment.
Dataset
Feedback data from IGN’s Game Wiki pages
Logical Approach
Using Excel and creating a pivot table based on categories I felt were useful within the dataset.
Methodology
- I started by visiting the URLs for each of the feedback to uncover some useful features and categories for grouping. It was only 24 sites, and from those visits, I built categories that lead to insightful filters for the data.
Categorizing the feedback based on the best judgment
- I’m not sure that was the most efficient way to scan webpages for content. I could have used an auto data scrape program, but live and learn.
- I created a pivot table to better visualize the subsections found within the feedback. I used categories such as missing or wrong information, and I filtered data based on what I thought was most important, like country code top-level domain (ccTLD).
#sentiment-analysis #data analysis #data analysis
