Raw data in. Decisions out.
My focus · Data analysis: BigQuery pipelines, churn analysis, Streamlit apps, report automation, GenAI integration, RFM segmentation.
Data & Business Analyst and Retention Program Supervisor. I take data projects from end to end: BigQuery pipelines, analysis and machine learning, through to dashboards and automation that cut reporting time by 50-70%.
// numbers from real work, full details further down
I work where data meets the teams that use it.
Today I lead the retention team at a fiber internet provider, covering churn, loyalty, and win-back. Before that I managed an outbound call team and reporting for automotive clients at OTO.com, and built a BI platform on Google Cloud for Rahsa Nusantara.
I work closely with CRM, Sales, Marketing, Product, and HR teams. My approach is always the same: find repetitive manual work, clean up the data, then replace it with a tool that runs on its own.
Every project on this page follows the same flow. These are the tools I use at each stage.
// scroll to move through the stages
Pulling data from exports, sheets, and APIs.
Storing it, scheduling syncs, and removing duplicates.
Cleaning, reconciling, and giving it meaning.
Serving it through apps, APIs, and dashboards.
Turning analysis into programs and decisions.
Here are some of them, from three companies, an internship program, and personal projects. Pick a category to filter.
A BigQuery data warehouse with daily to yearly data marts, a FastAPI backend on Cloud Run, and a React/TypeScript dashboard. A GenAI layer with SQL guardrails lets stakeholders ask about sales in plain language.
Tracks customers, revenue, ARPU, and bandwidth through a schema-driven calculation engine. Every cell is reconciled to the source Excel files, then synced to Google Sheets with a single command.
The team only uploads new connect, reconnect, and disconnect data. Customer segmentation, calculations, and reconciliation run on their own, with an AI narrative for every chart.
Analysis of unsubscribed accounts showed churn is driven more by billing-related disconnections than by customers' own decisions. The outcome: six prioritized retention program proposals.
Redesigned the member-get-member program (referral credit wallet, band-based rewards, reward caps) and a three-tier loyalty program with a daily customer health score to prevent churn.
Daily call outcome and agent KPI reports for Toyota Astra Finance, Hyundai, Geely, and Wuling campaigns, now built by an app straight from dialer CSV exports.
Matched leads and vehicle orders (SPK) against the client's dashboard figures. The result was a multi-sheet Excel workbook that resolved gaps between internal and client reports.
Lead rate dropped while connect rate held steady. By reviewing call recordings and data, the cause was traced to lead list quality, not agent or dialer performance.
Each lead is matched to the nearest dealer using Haversine distance in Python. Leads with no dealer in range are flagged for review.
The team fills in one sheet. Apps Script routes leads based in Java to the client's sheet and keeps the rest in the internal sheet.
The SOV report, once compiled by hand, now runs automatically, so analysts have time for urgent client requests.
Grew from a Streamlit prototype into a production Next.js app with Firestore, Firebase Auth, and Secret Manager. It includes 360° reviews with AI summaries and a validated HR data seeding pipeline.
Attendance and daily sales recaps for in-store promoters (SPG), sent straight to WhatsApp and run automatically by a cloud scheduler.
Standardized city names to the City/Regency format, converted the data to Parquet for BigQuery, and matched offline event attendees by phone number.
An internal AI assistant that replaced paid third-party tool subscriptions, cutting costs and speeding up the team's daily work.
Lexicon-based sentiment analysis of product reviews on Tokopedia and Shopee, packaged as a web app so the CRM team can run it themselves. Plus a customer churn prediction model.
Bundle recommendations based on purchase patterns, so bundles aren't decided by gut feeling. Plus a stock screener that combines PEG, Graham Number, Earnings Yield, ROE, and DER.
Two Streamlit apps built on Google Sheets: a cross-company project tracker that replaces Trello, and a stock report for a light steel store.
Two dashboards I designed from public datasets. Both follow the same flow: what happened, why, and who needs attention first. Tap an image to enlarge it.
public dataset · 2,058 unique customers · 3 pages
This dashboard ends in a customer list ready to export to the campaign team, not just charts.
public dataset · 10,000 customers · 3 pages · bilingual
The risk score is rule-based so the team can check its reasoning, and gender is deliberately left out of the score.
PT Indonesia Broadband Communication
PT Carbay Services Indonesia (OTO.com)
PT Bhinneka Rahsa Nusantara
Kampus Merdeka Sanggar Rahsa x Magang Merdeka MSIB
Send a message through this form. I usually reply within 1 to 2 business days (Jakarta time, GMT+7).
I'm building two sites of my own: one to sell products, one to share how I work and think.