Listening to the customer

The context

Consumer reviews can reveal what people value and where they encounter problems, but free-form text is difficult to compare. The project aimed to turn that feedback into an understandable analysis for brand decision-making.

How I worked

• Prepare and explore. Cleaned the review data with Python and Pandas, then used exploratory analysis and visualizations to understand the sample before interpreting the text.
• Examine sentiment. Applied NLP and sentiment analysis, using Gemini to help interpret consumer feedback and recurring themes.
• Build a coherent recommendation. Combined the cleaning, exploratory and AI-assisted analysis stages into a capstone notebook, linking recommendations to the review evidence.

Deliverables & takeaways

I produced visualizations and evidence-based recommendations, with the cleaning, EDA, AI analysis and capstone stages available in the repository. The preview shows the composition of the review sample, helping readers understand the basis of the comparison.

The project developed my ability to move from unstructured customer feedback to a documented analytical argument, while separating what the sample suggests from what it can establish.

Scope & limitations

The review sample is not a representative survey of all customers. Unequal sample sizes and model-generated sentiment can affect comparisons. The recommendations are analytical proposals; I did not measure their adoption or a subsequent business outcome.