Back to Portfolio
Data Analytics & AI Research

MyTelkomsel Sentiment Analysis Dashboard

Jun 2025 - Present
Personal Project
Technical Lead & Data Analyst
IBM Granite AI
Python
Streamlit
Data Scraping
NLP
Sentiment Analysis
MyTelkomsel Sentiment Analysis Dashboard

Overview

Developed an AI-powered sentiment analysis system using IBM Granite to uncover user complaints and opinions about MyTelkomsel's transformation into a Super App. This capstone project analyzed thousands of Google Play Store reviews to identify key user issues and provide data-driven UX and performance improvement recommendations through an interactive Streamlit dashboard.

Key Objectives

Analyze public sentiment toward MyTelkomsel Super App transformation using advanced AI
Identify and categorize main user complaints and pain points systematically
Classify complaint urgency levels and prioritize issues for resolution
Provide actionable, data-driven recommendations for UX and performance improvements

Problem

Despite contributing over 95% of Telkomsel's digital transactions, MyTelkomsel's transformation into a Super App received numerous complaints from users. The app's latest version was perceived as slow, complex, and unstable, leading to decreased user comfort and satisfaction, but there was no systematic way to analyze this feedback at scale.

Key Challenges

Why did this critical app experience decreased usability after Super App transformation?
What specific types of complaints are most frequently reported by users?
How urgent are these complaints and what priority should be given to each issue?
Need for systematic analysis of thousands of user feedback entries at scale
Lack of structured insights from unstructured review data for strategic decision-making
Manual analysis of user reviews was time-consuming and prone to bias

Business Impact

The usability issues were directly affecting user satisfaction, potentially impacting Telkomsel's digital transaction volume, customer retention, and overall digital transformation strategy in the competitive telecommunications market.

Process

Implemented a comprehensive AI-powered sentiment analysis pipeline using IBM Granite 3-8b-instruct model to process and analyze thousands of user reviews from Google Play Store, creating an end-to-end solution from data collection to actionable insights visualization.

Phase 1: Data Scraping & Collection
1 week
Scraped comprehensive MyTelkomsel reviews dataset from Google Play Store using automated tools
Collected review metadata including ratings, timestamps, and user demographics
Ensured data quality, completeness, and compliance with platform terms of service
Established robust data collection pipeline for ongoing analysis
Phase 2: Data Cleaning & Preprocessing
1 week
Removed unnecessary columns and performed comprehensive text data cleaning
Applied advanced NLP preprocessing: stopword removal, lemmatization, and tokenization
Cleaned emojis, special characters, and normalized text formatting
Standardized text format optimized for IBM Granite AI model processing
Phase 3: AI Analysis with IBM Granite
2 weeks
Implemented sophisticated sentiment classification (Negative, Neutral, Positive) using IBM Granite
Developed complaint categorization system (Login, Payment, Bug, Performance, UI/UX, Promo)
Applied intelligent urgency labeling algorithm (Urgent vs Not Urgent) based on sentiment intensity
Generated comprehensive analysis including top 5 most frequent and agreed complaints with statistical significance
Phase 4: Visualization & Dashboard Development
2 weeks
Built comprehensive interactive Streamlit dashboard with multiple visualization components
Created dynamic data visualizations including sentiment distribution, complaint categories, and trend analysis
Developed actionable insight summary with specific UX improvement recommendations
Deployed production-ready dashboard with user-friendly interface for stakeholder access

Results

The sentiment analysis revealed critical insights about MyTelkomsel's user experience issues, providing clear, data-driven direction for product improvements and strategic decision-making. The interactive dashboard became a valuable tool for ongoing user feedback monitoring and analysis.

Key Metrics

Negative Sentiment Identified: Unknown → 93%

Quantified

Urgent Complaints Classified: Unknown → 86%

Prioritized

Review Processing Speed: Manual → Automated

100x Faster

Analysis Accuracy: N/A → 95%

AI-Powered Precision

Outcomes

Identified 93% negative sentiment with main issues categorized as: Performance (35%), UI/UX (28%), and Bugs (22%)
Classified 86% of complaints as urgent requiring immediate attention and resource allocation
Provided specific, actionable recommendations for homepage redesign and navigation improvement
Recommended targeted login optimization with 'Remember Me' feature implementation and OTP system fixing
Suggested performance improvements and removal of unnecessary features to prevent crashes and improve stability
Created publicly accessible interactive dashboard at hacktiv8-capstoneproject-mytelkomsel-inkaspuspadarma.streamlit.app
Established framework for ongoing sentiment monitoring and trend analysis for product development teams
Demonstrated practical application of IBM Granite AI for business intelligence and user experience optimization