A chain of department stores receives a myriad of Google reviews for their various locations. The challenge lies in efficiently extracting and interpreting actionable insights from this substantial data. An ideal solution would not only facilitate understanding of customer sentiment for each individual store, but also offer a summarized view across all locations. Further, the store is interested in expanding this analysis to include their competitors. Can we devise a way to understand customer sentiments about competitors' stores from these online reviews as well?

To tackle this challenge, we employed several Natural Language Processing (NLP) techniques. Our strategy entailed the development of a centralized platform that aggregates, processes, and visualizes Google review data for both our stores and our competitors. We began by extracting reviews for each store location from Google reviews, including competitor stores. Next, we applied sentiment analysis to categorize these reviews into positive, negative, and neutral sentiments and topic modeling to identify key themes discussed in the reviews, allowing us to pinpoint specific areas of praise or concern.

This process resulted in the creation of a comprehensive dashboard that not only provides a snapshot view of customer sentiment for each store, but also highlights trending topics across all locations. By extending this analysis to competitors' reviews, we could gather industry insights and benchmark our performance against them.

This holistic view of customer sentiment offers valuable insights into strengths and weaknesses and allows for informed, strategic decision-making for business improvement. It also gives us a competitive edge by providing insights into our competitors' customer sentiment, helping us identify industry trends and opportunities for differentiation.

‍

We utilized Python, alongside libraries such as Beautiful Soup and NLTK for web scraping and data cleaning. Sentiment from the review was extracted with a lexicon and rule-based analysis with Vader and a LDA model was used for topic modeling. The insights derived from the sentiment analysis and topic modeling were consolidated into an interactive dashboard using Tableau. This dashboard provided an overall view of the sentiment across all stores and highlighted key discussion topics from the reviews.

The entire process was replicated for competitor reviews to offer a comparative understanding of customer sentiments and prevalent themes.

By employing this method, we effectively turned unstructured online reviews into structured, actionable insights for our stores and our competitive analysis.

‍