Review Trend Analysis
Surface product issues hiding in the review feed before they become returns.
Problem
Negative patterns in reviews, like quality complaints or the same product issue coming up over and over, get lost when there are too many reviews to read. By the time anyone noticed, it had usually already turned into higher returns.
Approach
I scraped Google reviews and the on-site reviews into one organized dataset, then had a custom GPT look across all of them for repeating themes, shifts in sentiment, and product-specific issues. The output is a short summary that flags which products are trending negative and why.
Result
Can help catch product issues early before they turn into return spikes. Gives the team a regular read on what customers are saying without anyone having to go through every review by hand.