What Are Customers Really Saying?
Customers already tell you exactly what's wrong — and what's right — in reviews, surveys, and support messages. The problem is volume: nobody has time to read thousands of reviews and remember the patterns. So the feedback piles up unused.
The goal here was to read every review automatically and surface the recurring themes — what people keep praising, what they keep complaining about. The data was 21,570 reviews of 8,327 Nike products, written by more than 20,000 reviewers.
The review text was cleaned and standardized, then run through topic modeling (LDA) to group it into distinct themes, and through sentiment analysis to score each review positive, neutral, or negative — cross-checked against the star ratings so the sentiment could be trusted.
The themes that surfaced were specific and actionable, not vague:
Each of those is a concrete instruction to a product or marketing team — fix the watch battery life, add pockets to the bags, advise customers to size up.
What this means for your business: any business with reviews, survey responses, or support tickets is sitting on the same goldmine — and the same problem of too much to read. This approach turns all of that unstructured text into a ranked list of what customers keep saying and how they feel about it. It tells you what to fix, what to promote, and what your customers actually value.