Real-Time Crisis Sentiment Analysis System
Client: A Major Telecom Provider
Project Goal: To develop a system that monitors, analyzes, and visualizes public sentiment in real-time, especially during crisis events, to identify specific customer pain points and discussion topics.
The Challenge: During high-pressure events, like the Istanbul earthquake, the client's network experienced massive strain, leading to service disruptions. This resulted in a surge of negative customer feedback on social media. The client needed a way to move beyond simple "positive/negative" metrics and understand the exact topics customers were discussing (e.g., "slow data," "call drops," "no signal") to prioritize responses and technical adjustments.
Our Solution: We designed and implemented an end-to-end sentiment analysis pipeline that scrapes public data, processes it using advanced AI, and visualizes the findings in an interactive dashboard.
Key Technologies & Components:
- make.com: Utilized as the primary automation engine to build scrapers and connect all system components.
- Custom Web Scraper: Actively monitored social media and web sources for brand mentions, collecting the content and available user data (like location).
- Central Database (Google Sheets/Airtable): Stored all scraped data, which was then systematically enriched with analysis results.
- AI Sentiment & Topic Analysis:
- Google Natural Language API: Employed for sophisticated sentiment analysis and entity/topic extraction.
- Amazon Comprehend: Used as an alternative or supplementary AI for robust analysis.
- OpenAI API: Integrated for nuanced understanding and categorization of topics.
- Looker Studio (Google Data Studio): Connected directly to the database to provide a real-time, interactive visualization dashboard.
- Automated Email Reporting: Scheduled workflows in make.com sent daily or weekly email summaries of the current sentiment landscape to supervisors.
How it Works (System Flow):
- Data Collection: A make.com workflow constantly scrapes the web for new brand mentions.
- Database Entry: All new comments and posts are saved as new entries in the database.
- AI Analysis: A second workflow is triggered, sending the new text to Google and Amazon's AI APIs.
- Data Enrichment: The AI returns the sentiment (positive, negative, neutral), the key topics discussed, and a "certainty score" (confidence level of the analysis). This information is then updated in the database.
- Live Visualization: The Looker Studio dashboard, accessible to the client's team, reflects this new data instantly. Supervisors can filter by date range (e.g., "last 24 hours," "last two months") to see the most talked-about positive, negative, and neutral topics.
- Automated Alerts: A scheduled workflow compiles key metrics and trends into a concise email report, notifying supervisors of the market's sentiment without them needing to manually check the dashboard.
Project Impact & Benefits:
- Real-Time Crisis Insights: The client could instantly see what customers were saying during a crisis, allowing for faster and more targeted public relations and technical responses.
- Specific Topic Identification: Moved beyond generic sentiment to pinpoint exact issues, enabling data-driven decision-making.
- Data-Driven Reporting: Automated dashboards and email alerts provided accessible insights to stakeholders at all levels.
- High-Profile Recognition: This innovative solution was successful and received industry recognition, being featured on the official website and LinkedIn a social media of one of our key technology partners (make.com).