Leading FMCG Company - Real-Time Sentiment on 500K+ Daily Mentions
How we built an NLP social listening tool that turned social conversation into product decisions in hours instead of weeks.
The numbers, in production.
A premium cookie brand in a leading FMCG company's portfolio had no systematic way to see what consumers were saying about it across social platforms.
Product strategy decisions ran on survey data that arrived weeks after the fact. By the time the team saw a shift in sentiment, the conversation had already moved on.
The solution
We built an NLP-powered listening tool that reads social conversation as it happens and puts the signal in front of the product team.
Classify sentiment with Transformer models
We used Transformer-based NLP models (BERT) in Python to classify the sentiment of every relevant post and comment.
Index social data in real time
Elasticsearch indexes the incoming social data streams, so the team can search and filter conversation as it happens.
Process events without servers to manage
AWS Lambda, API Gateway and S3 handle serverless, event-driven processing that scales with conversation volume.
Put it on one dashboard
A React dashboard gives the product strategy team a live view of sentiment, so they can act on it the same day.
- Python (Transformers)
- NLP / BERT
- Elasticsearch
- AWS Lambda
- API Gateway
- S3
- React
The result
The brand now has a real-time sentiment dashboard tracking more than 500,000 social mentions a day, and the sentiment model classifies them with 89% accuracy.
The bigger change is speed. The product strategy team moved from lagging survey reports to live signals, and cut the time from insight to decision from weeks to hours.
- 500K+ social mentions tracked every day
- 89% sentiment classification accuracy
- Insight-to-decision time cut from weeks to hours
- Serverless architecture that scales with conversation volume
Survey data tells a brand what consumers thought last month. Real-time listening tells it what they think today, and that is what a product team can act on.