The client consists of a team of health-conscious entrepreneurs, doctors, and professional chefs who make pre-cooked, nutritious, and tasty food available to their consumers. The subscription model simplifies consumers’ wellness goals and also meets their nutritional requirements with microwavable food packages cooked per their needs.
The client solicits regular consumer feedback and comments on their website and mobile application to improve the quality of their services. Previously, to infer consumer insights, the feedback was ingested into a system to tag, sort, and classify them into predefined custom categories.
However, this system was not intuitive enough to introduce new categories based on the ever-evolving feedback patterns. Also, the tool was proving to be cost prohibitive and needed increased manual efforts to understand and analyze the feedback.
Machine Learning leads the way to better understand consumer feedback
To reduce the manual efforts and eventually automate content classification and analysis, the client worked with their digital consulting partner, Persistent. Using AutoML NLP, Persistent trained two machine learning models on the client’s consumer feedback data stored in Google BigQuery.
The first model is a text classifier that automatically classifies the feedback into one of the well-defined, high-level, and broad hierarchy of classes defined by the client. The second model measures the sentiment score of the feedback comment on a continuous scale, in contrast to the binary scale used previously.
At the end of this data pipeline, the model displays the results on a dynamic dashboard. Also, this serverless and scalable solution allows the client to introduce new categories as required.
The solution, which was delivered end-to-end in just six weeks, helps the client to operate at a significantly lower cost, achieving as much as 50% cost savings as compared to the previous solution.