Sentiment Analysis Case Study: Analyzing 7,000 Tokopedia Customer Tweets
- admin
- Apr 21
- 5 min read
Updated: 1 day ago
Tokopedia's official support account, @TokopediaCare, handles over 1 million posts on X (formerly Twitter) alone. Customers raise complaints about everything from failed top-ups and cashback errors to delayed refunds and unresponsive support. At that volume, reading and categorizing each tweet manually is not a viable option.
This case study walks through how BI Solusi built an AI-powered sentiment analysis system that helped Tokopedia extract structured insights from 7,000 customer tweets, identify top monthly issues, track recurring problems, and detect viral complaints before they escalate.
The Problem with Monitoring Customer Sentiment at Scale
Customers tweet complaints in real time, and each tweet can contain multiple issues at once. One message might express frustration with a failed promo while also criticizing response times from support.
Monitoring this manually creates two problems.
First, it is too slow. By the time a team reads, categorizes, and escalates a complaint, the issue may have already spread.
Second, manual review misses the pattern. A single tweet about a failed top-up is a support ticket. Fifty tweets about the same issue in one week is a systemic problem that needs an operational fix.
Why Tokopedia Needed Aspect-Based Sentiment Analysis
Standard sentiment analysis tools classify a piece of text as positive, negative, or neutral. For social media monitoring at an e-commerce platform, that output is too simplified to be useful.
Consider a tweet like: "The new features are crazy, but sometimes the service is lacking." A traditional model reads the full sentence and returns a single label — most likely negative, because of the word "lacking." The positive sentiment about the features disappears entirely.
Aspect-based sentiment analysis identifies the distinct topics within a tweet and assigns a sentiment to each one. For the same tweet, the output would be:
Features: Positive
Service: Negative
If a company sees a spike in negative sentiment, the next question is always: negative about what, specifically? Aspect-based analysis answers that without requiring a human to read every post.
For Tokopedia, where customers regularly raise multiple complaints in a single tweet, this level of granularity was the only approach that would produce actionable data.
How AI Sentiment Analysis Works on Social Media Data
The system BI Solusi built as part of its NLP and sentiment analysis services processes each tweet through a pipeline that handles both classification and summarization.
Data collection: Tweets directed at or mentioning Tokopedia are pulled from X. For this project, the dataset covered 7,000 customer tweets.
Topic extraction: Each tweet is analyzed to identify which topics or service areas it references. For example, top-up, refund, delivery, payment, registration, or customer support.
Sentiment classification per topic: For each identified topic, the model assigns a sentiment: positive, negative, or neutral. Sentiment is attached to a specific topic, not to the full text.
Issue summarization: The system aggregates results by topic across a given time period and generates a summary of what customers are saying, removing the need to read individual posts.
Trend tracking over time: The system plots issue volume over time, making it possible to see whether a specific issue is growing, stable, or declining across weeks and months.
Findings from 7,000 Tokopedia Customer Tweets
The analysis surfaced three types of insights: monthly issue summaries, recurring complaint patterns, and early viral trend detection.
June 2023: Top Issues Identified
The Issues Summary dashboard for June 2023 surfaced two dominant complaints:
Customer Support: Slow response times and unsatisfactory complaint handling
Payment and Finance: Delays in e-money balance updates and difficulties withdrawing funds as a seller
Recurring Issues: Top-Up Appears Month After Month
The Issues Tracker shows popular issues plotted over time. Across the tracking period through November 2023, top-up issues appeared as the dominant complaint in nearly every month. Delivery and refund issues also recurred consistently.
A recurring issue in the data signals that the underlying problem has not been resolved — and the tracker makes that visible without anyone needing to manually compare month-over-month reports.
September 2023: A Viral Complaint Detected Early
The Trend Analysis dashboard for September 2023 flagged a sharp spike in a specific complaint: customers reporting that they ordered an iPhone but received a rock instead. The spike peaked in mid-September and was visible before it became widely discussed.
Early detection gives a company a narrow window to respond and investigate before the story spreads further.
What E-Commerce Brands Can Learn from This Approach
Social media sentiment data becomes significantly more useful when it is structured, not just aggregated. A dashboard showing 60% negative sentiment is not actionable. A dashboard showing that top-up complaints appeared in the top issues every month for six consecutive months — with a summary of what customers specifically experienced — is.
Three takeaways that apply to any e-commerce brand handling high-volume customer interactions:
Topic-level classification produces decisions, not just reports. Knowing which service areas drive negative sentiment lets teams prioritize fixes at the right level.
Recurring issue tracking reduces repeat escalations. If the same problem surfaces every month, it is an operations or product issue — not a support queue problem.
Trend detection only works if it is fast enough to act on. A report produced two weeks after a viral complaint has peaked has limited value. Automated monitoring closes that gap.
Build a Customer Sentiment Monitoring System with BI Solusi
Tracking customer sentiment at scale is only useful when the output is specific enough to act on — and that requires the right system design from the start. With BI Solusi, you get more than a sentiment analysis tool. Our certified consultants work with you end-to-end: from defining the right approach for your data sources, to building the classification pipeline, to delivering dashboards your team can act on.
Talk to our data science team to see how we can build this for your platform.
BI Solusi is your trusted partner for data-driven success in Indonesia, serving companies in the Southeast Asia region and beyond. We specialize in implementing cutting-edge Data Analytics, Business Intelligence platform, and Big Data solution, complemented by expert Data Science services.
We offer flexible nearshore and offshore BI implementation models to meet your specific needs and deliver the highest-quality results.
Our BI Consulting expertise encompasses Data Integration services (ETL), Data Warehousing, and the utilization of Data Visualization tools such as Microsoft Power BI, Qlik Sense, and Tableau for Reports and Dashboards implementation.
Let us help you unlock the full potential of your data and achieve your business goals.
After participating in the kroger feedback survey conducted by the the retailing company, Customers can chance to obtain kroger fuel points