Visual summary of operating lessons from Crystal Widjaja.

Lessons from Crystal Widjaja

Crystal Widjaja was Gojek's first technical data hire. She says the teams she built helped the company grow from 30,000 to more than 10 million transactions a day. Her roles there spanned business intelligence, growth, and Chief of Staff; she later advised startups and co-founded Generation Girl to expand young women's access to STEM. — Crystal Widjaja — About.

Part 1: Data Infrastructure & Engineering

  1. On Data Integrity: Establish what the numbers actually measure before relying on sophisticated analysis. — Why Most Analytics Efforts Fail.
  2. On Garbage In, Garbage Out: Show engineers how inaccurate source data obscures the impact of their work, so data quality becomes a shared concern. — Tech Lead Journal #13.
  3. On Database Scaling: Gojek moved from manual reports and MySQL/MongoDB inputs to Postgres, then to a BigQuery data lake as scale demanded; infrastructure evolved with the queries the business needed. — Tech Lead Journal #13.
  4. On Democratizing Data: Design self-service analytics around the questions that product, marketing, and operations teams need answered, even when those teams do not write SQL. — Why Most Analytics Efforts Fail.
  5. On Analytical Bottlenecks: Automate repeated manual reporting so the data team and business users can spend more time on decisions rather than waiting for routine exports. — Tech Lead Journal #13.
  6. On Building Trust: Use clear event definitions and reliable pipelines so colleagues can understand and trust the data they use. — Why Most Analytics Efforts Fail.
  7. On Tool Empowerment: Giving customer-care colleagues self-service access let some teach themselves SQL and answer operational questions directly. — MALAKA — Crystal Widjaja on Data.
  8. On Problem-First Architecture: Start with the decisions and user questions the data system must support, then automate and upgrade infrastructure as needs become clear. — Why Most Analytics Efforts Fail.
  9. On Defining Metrics: Agree on what each tracked event means before teams use the same metric to make different decisions. — Why Most Analytics Efforts Fail.

Part 2: Product Management & Alignment

  1. On Altitude Maps: Map product decisions and input levers to a manager's and company's higher-level outcome metrics, then revisit the map as you learn. — Stop Data Theater: Use Altitude Maps.
  2. On Unmeasurable Metrics: When an outcome is difficult to measure directly, use a practical proxy and state its limits instead of pretending the data is perfect. — Reforge — Unsolicited Feedback E9.
  3. On Managerial Alignment: Connect your product measures and day-to-day levers to the broader outcomes your manager owns. — Stop Data Theater: Use Altitude Maps.
  4. On Strategic Decisions: Spend analytical effort on choosing the right problem, then run quick experiments that can reveal whether the hypothesis is wrong. — Crystal Widjaja on LinkedIn.
  5. On Feature Bloat: Gojek shut down cleaning, massage and beauty services when they confused customers and did not fit the strengths of its driver platform. — Martech Family Interview.
  6. On Product Speculation: Test the intended experience cheaply, as Gojek did with driver-assisted subscription pitches and simple screen overlays, before building a full feature. — Lenny's Podcast.
  7. On User Feedback: Use observed user behavior to test a product hypothesis, rather than assuming an untested feature request will solve the problem. — Lenny's Podcast.

Part 3: Growth & Experimentation

  1. On Scrappy Growth: Use cheap, concrete tests to learn what users will do before investing in heavy growth infrastructure. — Lenny's Podcast.
  2. On Frequent Experiments: Run focused experiments often enough to replace assumptions with observations; the useful cadence depends on the problem, not an arbitrary daily quota. — Lenny's Podcast.
  3. On Interpreting Patterns: A behavior correlated with retention is a clue to investigate, not proof that changing that behavior alone will cause retention. — Tech Lead Journal #13.
  4. On Marketplace Supply: Account for the driver side of a marketplace and the channels through which supply reaches customers before treating acquisition spend as the whole growth plan. — Lenny's Podcast.
  5. On Growth Teams: Give an early growth hire a known bottleneck to solve and test for statistical and experimental judgment, not just familiarity with growth tools. — Lenny's Podcast.
  6. On Incentive Design: Gojek used incentives to help drivers introduce GoPay to first-time users, showing how marketplace participants can support adoption. — Lenny's Podcast.
  7. On Retention Over Acquisition: Match promotions to the product's natural use frequency; too many messages can drive unsubscribes and costly reacquisition. — Tech Lead Journal #13.
  8. On High-Leverage Actions: Look for the largest current constraint in the user journey and test an inexpensive intervention there before building a comprehensive solution. — Martech Family Interview.
  9. On Failing Experiments: When a test fails to generalize, examine its design and rollout context before treating the result as a permanent verdict on the idea. — MALAKA — Crystal Widjaja on Data.

Part 4: AI & Context Engineering

  1. On Agent Workflows: A product manager can make AI more useful by defining a repeatable workflow with clear inputs and completion checks, rather than collecting isolated prompts. — Semi-Technical — Chief of Stuff.
  2. On Context Engineering: Maintain relevant task context and explicit acceptance criteria so an AI workflow has enough information to complete the right job. — Semi-Technical — Chief of Stuff.
  3. On AI as Chief of Staff: Delegate routine follow-ups, drafts and note triage to AI while keeping novel relationship work and judgment with a person. — Semi-Technical — Chief of Stuff.
  4. On Command-Line Literacy: Product managers who learn basic command-line workflows can use current AI tools more directly and inspect what those tools do. — Atlassian — PMs and the CLI.
  5. On Productivity Traps: Use AI to finish low-leverage work rather than creating more drafts and material for yourself to review. — Atlassian — PMs and the CLI.
  6. On Human Evaluation: Human judgment remains necessary for novel, high-context choices even when AI can draft or organize routine work. — Semi-Technical — Chief of Stuff.
  7. On the AI Era: Define the outcome and review AI-produced work instead of assuming generated code or text is correct by default. — Semi-Technical — Chief of Stuff.
  8. On Automating Data Pulls: Validate AI-written SQL against the actual schema before trusting a natural-language data pull. — Semi-Technical — Why LLMs Are Bad at SQL.
  9. On Technical Intuition: Understanding basics such as client-side and server-side behavior helps a product manager spot when an AI prototype cannot work in production. — MALAKA — Crystal Widjaja on Data.
  10. On Iterative Prompting: Iterate on workflow, context and testable completion conditions; better prompts alone are not enough. — Semi-Technical — Chief of Stuff.

Part 5: Women in STEM & Education

  1. On Early Exposure: Give girls an early chance to try coding and technology, before stereotypes narrow the range of careers they consider. — KRASIA — Crystal Widjaja Interview.
  2. On Mentorship: Pair girls with university-age mentors and women in technical roles so they can see accessible paths into STEM. — Tech Lead Journal #13.
  3. On Building Confidence: Generation Girl gives participants a safe place to try technical work, make mistakes and discover whether they enjoy it. — Tech Lead Journal #13.
  4. On Bridging the Gap: Address unequal access to technical opportunities before girls reach the point of choosing a job or degree. — KRASIA — Crystal Widjaja Interview.
  5. On Normalizing STEM: Make it ordinary for girls and women to see themselves trying and working in technical fields. — Tech Lead Journal #13.
  6. On Safe Environments: Use free holiday bootcamps as welcoming places where girls can experiment with STEM alongside peers and mentors. — KRASIA — Crystal Widjaja Interview.
  7. On Future Leaders: Early technical education and mentorship can widen the path for more women to pursue technical careers and leadership. — KRASIA — Crystal Widjaja Interview.
  8. On Giving Back: Invite working technologists and partner companies to mentor girls and expand access to STEM programs. — KRASIA — Crystal Widjaja Interview.

Part 6: Career Trajectory & Angel Investing

  1. On Starting at the Bottom: She began at Gojek by asking what the reports actually measured and how reliable the figures were before helping build a broader data organization. — BRAVE E92 — Crystal Widjaja.
  2. On Parallel Lives: A career can connect roles in product, investing, advising and nonprofit work rather than forcing them into separate identities. — CreativeMornings — Crystal Widjaja on Parallel.
  3. On Angel Investing: Her investment in Eppo drew on firsthand experience with difficult marketplace experiments and a belief that better tooling could make rigorous testing easier for teams. — Martech Family Interview.
  4. On Career Progression: Her move from analytics and growth into Chief of Staff work changed the problem from individual analysis to company-wide management systems. — BRAVE E92 — Crystal Widjaja.
  5. On Choosing Companies: She examined Gojek's underlying user problem and market before joining; a career move should be judged by the problem the team is solving. — MALAKA — Crystal Widjaja on Data.
  6. On Chief of Staff Roles: A Chief of Staff can identify neglected operating problems and build practical management support across an expanding company. — BRAVE E92 — Crystal Widjaja.
  7. On Leaving Comfort Zones: She left a comfortable Gojek role to test whether her skills would transfer and to keep learning in a different setting. — Tech Lead Journal #13.

Part 7: Leadership & Organizational Debt

  1. On Scaling Teams: As Gojek grew, management training and shared operating practices became necessary alongside growth in the data and product teams. — BRAVE E92 — Crystal Widjaja.
  2. On Organizational Debt: Fast growth leaves gaps in management and coordination; repair the debt that causes burnout while allowing some flexibility that helps a changing company adapt. — BRAVE E92 — Crystal Widjaja.
  3. On Hiring Scrappily: Test an early growth hire's grasp of statistics, random sampling and first-principles experiment design, not just prior titles or tools. — Lenny's Podcast.
  4. On Information Silos: Shared event definitions and discoverable data documentation make it easier for teams to understand and use the same information. — Why Most Analytics Efforts Fail.
  5. On Managing Up: At Kumu, she asked what visibility her manager lacked and tried to turn new information into useful context and insights. — Your Activation Experience — Crystal Widjaja.

Part 8: Philosophy & Personal Development

  1. On Rationality: Focus discussion on risks that warrant action or awareness, rather than spending scarce attention debating probabilities that will not change a decision. — Reforge — Unsolicited Feedback E9.
  2. On Productivity Systems: Adapt a productivity workflow through repeated use and refactoring until it reliably handles the routine work you actually have. — Semi-Technical — Chief of Stuff.
  3. On Embracing Failure: She values the mistakes she made at Gojek because they taught her what to change; a failed attempt can be useful evidence. — BRAVE E92 — Crystal Widjaja.
  4. On Burnout Risk: After experiencing long hours and burnout risk, she weighed the personal cost of rebuilding another organization before taking on the work. — BRAVE E92 — Crystal Widjaja.
  5. On Evaluating Advice: For messy organizational problems, she sought experienced perspectives because generic online answers missed the context and tradeoffs. — BRAVE E92 — Crystal Widjaja.
  6. On Deep Work: She schedules three-hour writing blocks three days a week, making time to draft or edit even when inspiration is absent. — Martech Family Interview.
  7. On Intentional Practice: She relies on intentional writing time rather than waiting for inspiration to arrive; consistent engagement helps work move forward. — Martech Family Interview.
  8. On Legacy: At Gojek, some of her most durable work was manager training, culture documentation and OKR scaffolding built for a larger organization. — BRAVE E92 — Crystal Widjaja.