T-Emotions

Inspiration

We've all been there—scrolling through customer feedback and wondering, "Is this an urgent issue or just someone venting?" When you're dealing with hundreds of comments across Reddit, Twitter, and review sites, it's nearly impossible to catch problems before they blow up. That's exactly what we wanted to solve with the Customer Happiness Index.

T-Mobile gets tons of customer feedback every day on Reddit alone. Some posts are people raving about their 5G speeds. Others are frustrated customers dealing with billing errors or network issues. The problem? There's no easy way to separate the signal from the noise in real-time. By the time someone manually reviews feedback, hours or even days have passed—and an angry customer might have already switched carriers.

Solution

We built an AI-powered dashboard that reads customer feedback and instantly tells you: Is this positive, neutral, or negative? Should we escalate this to support, or celebrate this win with the team?

Here's how it works:

Data Collection — We scraped 100 Reddit posts from the T-Mobile community using Python and the Reddit API (PRAW). Real talk: Reddit users don't hold back, so it's perfect for honest feedback.
Cleaning the Mess — Raw Reddit posts are... messy. Typos, emojis, URLs, markdown formatting everywhere. We built a 7-stage cleaning pipeline that strips out the noise, removes duplicates, and filters out anything that's too short or not in English. Started with 100 posts, ended with 86 high-quality ones.
Teaching the AI — We used TextBlob to label the sentiment of each post, then trained two machine learning models (Logistic Regression and Random Forest) to predict sentiment on new feedback. We picked the best performer and saved it for deployment.
Building the Dashboard — This is where it gets fun. We created a 5-page Streamlit dashboard that looks and feels like a T-Mobile app (custom Magenta and Black colors, smooth animations, the whole deal). It's not just pretty—it's functional:

Features

Dashboard Overview: Shows your "Happiness Index" (a weighted score combining all sentiment), live metrics, and instant alerts for both urgent issues and customer wins. Historical Analytics: Dive deep into the data—what keywords show up in negative vs positive feedback? What are people actually talking about? Live Sentiment Detector: Paste any customer comment and get instant predictions. "This is negative—escalate to support." "This is positive—share it with the team!" Issue Detection: Automatically categorizes problems into Network, Billing, Customer Service, or Device issues. Shows you exactly where things are going wrong. Moments of Delight: Because it's not all bad news! This page highlights your best customer testimonials so you can celebrate wins and use them for marketing.

What We Discovered

The data told some interesting stories. 54.7% of T-Mobile Reddit feedback is positive (people love the 5G speeds and pricing!), but there are clear pain points:

eSIM transfers are a nightmare for customers (multiple posts about failed transfers and long wait times) Billing errors create major frustration 5G UC performance is inconsistent in some areas

These insights came from just 86 posts. Imagine what you could learn from thousands of comments analyzed in real-time.

Challenges we ran into

Accomplishments that we're proud of

The Impact

Right now, this dashboard analyzes historical data. But here's the vision: Connect it to live Reddit streams, Twitter feeds, and customer support tickets. Every time someone posts a complaint, the system flags it within minutes instead of hours. Support teams get automatic alerts. Product teams see which features customers love (or hate). Marketing gets authentic testimonials delivered on a silver platter. Bottom line: Faster response times, happier customers, less churn. And it all started with 86 Reddit posts and some Python code.

What's Next for T-Emotions?

We're not done. The keyword extraction needs work (it's pulling generic terms like "phone" and "mobile" instead of sentiment-specific words). We want to implement TF-IDF-based keyword analysis and add bigram/trigram support so we can catch phrases like "customer service" or "billing issue." We also want to scale this up—connect to live APIs, expand to other carriers, maybe even predict churn risk based on sentiment trends. The foundation is solid; now it's time to build on it. If you're a telecom company (or any business drowning in customer feedback), this is your early warning system. Catch problems before they explode. Celebrate wins while they're still fresh. Keep your customers happy. Because at the end of the day, happy customers don't leave.