Challenge
Marketing improves only when the customer feedback and sentiment is interpreted correctly. It was extremely difficult to analysis large amount of textual feedback and filter out actionable takeaways. Due to time constraints, It is never feasible for marketers and sellers to read and scan feedback for each row or stakeholder. While the data was being collected, stored and lying untamed in the database for years, there was no solution available to aggregate it and provide a cockpit view to analyze customer sentiments by area, product, event or time. The objective of this dashboard was to view, analyze and interpret the free text feedback captured post event and provide insights around the experiences and sentiments. While these responses were available in the database on per event basis, there is no platform available to view. analyze and interpret the sentiments and experiences. This project aimed to provide a platform to achieve the same
Idea and Solution
Using the power of Azure Machine Learning AML to process textual comments and the capabilities of Microsoft Power BI to transform and visualize, I created a Power BI solution where we classified feedback into positive, neutral or negative sentiments and , it becomes super easy to filter by any program or event date and view the programs with best positive or negative sentiments. This dashboard provides information based on Azure machine learning AML algorithm sentiment analysis and classifies each descriptive response into:
1. Positive
2. Neutral
3. Negative sentiments
GitHub: Sentiment Analysis using Azure Text Analytics & Power BI
Results
This dashboard is a gamechanger when it comes to getting real time customer insights and being able to action them. As we get enormous amount of unstructured textual data, this helps in understanding the sentiments and experiences for respective event/programs. Further, a custom logic on the top of Azure ML algorithm ensures that the model is trained and retrained basis the feedback from stakeholders. Being a predictive AI model, this may not be 100% accurate representation, but has been highly effective to identify the overall sentiment and to improve decision making #responsibleAI #transparency
Recent Comments