Leveraging Machine Learning in Customer Sentiment Analysis

Posted by | Shashank Kulshrestha

In a highly competitive market, staying ahead of customer expectations is critical for any manufacturer.

A leading US plumbing fixture manufacturer wanted to develop a groundbreaking shower filter capable of filtering and revitalizing water. To ensure the success of their new product, the manufacturer’s innovation team sought to learn from the mistakes of similar product manufacturers and develop a solution that met customer expectations, using customer sentiment analysis.

Objective:

The company wanted to guide the development of their new shower filter by using machine learning to collect and analyze the vast pool of publicly available competitors’ customer product reviews to deliver customer sentiment analysis. This approach aimed to provide valuable insights, pinpoint product shortcomings and foster a comprehensive understanding of customer preferences.

Solution:

To help them achieve their objective, Fuld implemented a comprehensive text analytics solution, leveraging cutting-edge techniques and algorithms:

1. Data Gathering: We developed a Python-based tool to gather the most recent 5,000 reviews from a major online marketplace. To ensure a comprehensive analysis, the data encompassed a span of three years.
2. Data Preprocessing: We parsed and cleansed the gathered data using Python, then applied preprocessing techniques to enhance the quality of the data for analysis. This included removing unnecessary punctuation, expanding contractions, lemmatizing words to their root form, and eliminating stop words.
3. Sentiment Analysis: We applied VADER and BERT, two powerful algorithms, to identify the sentiment expressed in individual reviews and enable the team to understand the overall sentiments of customers towards competitors’ products.
4. Topic Modeling: The Latent Dirichlet Allocation (LDA) algorithm identified the dominant topics within customer reviews and helped uncover the key themes discussed by customers, providing valuable insights into their preferences and concerns.
5. Sentiment Analysis by Topic: In addition to identifying topics, we performed a sentiment analysis on each topic. This enabled the team to understand the sentiment associated with each topic and its impact on customer perception.
6. Keyword Identification: Fuld developed a Machine Learning (ML) model to identify the keywords that contributed most towards 5-star reviews. This step played a critical role in identifying the factors that held the highest value for customers, guiding the prioritization process during the product development phase.

Outcome:

The implementation of this comprehensive text analytics solution yielded several significant outputs:

1. Interactive Dashboard: We created a user-friendly dashboard in Tableau, enabling users to navigate the results dynamically. It provides the organization with a visually appealing and intuitive way to explore the collected data and gain actionable insights.

1. Key Learnings: The analysis uncovered essential learnings for the innovation team. Critical factors influencing customer satisfaction such as build quality, performance, long-term usage, value for money, perception and customer service emerged as prominent themes which reqiure special attention during the product development process.
2. Top Keywords: Through the identification of the top 20 keywords that contributed most to delivering 5-star ratings, the innovation team acquired valuable guidance on the areas requiring focus. These keywords acted as a compass, directing their efforts towards meeting customer expectations and delivering an exceptional product.

Conclusion:

By leveraging advanced text analytics techniques, the US plumbing fixture manufacturer was able to successfully collect and analyze competitor product reviews. The insights derived from the analysis enabled its Innovation Team to make informed decisions and guide the development of their new shower filter. The Product Development team is now considering these findings and developing a new product to better address customer needs.

This case study highlights the power of text analytics to help businesses understand customer sentiment, identify product shortcomings, and ultimately drive successful product innovation within a B2B context.