Lenny's PodcastHow to scrappily hire for, measure, and unlock growth | Crystal Widjaja, Gojek and Kumu
CHAPTERS
- 0:00 – 7:55
Crystal’s unconventional path into product and data leadership
Crystal shares how she went from a poli-sci major and investment banking research into building databases, then intentionally pivoting into tech. She explains how cold-emailing in Southeast Asia led her to Gojek, where she gradually took on data, risk, marketing, and ultimately growth.
- •First-generation career navigation: finding a job via Craigslist, landing in investment banking research
- •Self-taught database building as an early “product instinct” moment
- •Pattern-matching opportunities in Southeast Asia and leveraging Indonesian roots
- •Cold-emailing Gojek and taking a high-uncertainty bet
- •Progression at Gojek: data → fraud/risk → performance marketing → growth
- 7:55 – 9:40
Why Gojek scaled into a “super app” (and what Kumu is building)
Lenny and Crystal set context on the scale of Gojek/GoTo and Crystal’s current role at Kumu. Crystal contrasts Gojek’s transactional super-app model with Kumu’s social “all-in-one” community and live interaction use cases.
- •Gojek/GoTo scale: 170M users and 20+ services across SEA
- •Super app value: multiple layers of services to fill market gaps
- •Gojek as transactional (rides/food/payments) vs. Kumu as social/community (audio/video/multi-seat)
- •Kumu’s geographic focus (Philippines) with top-grossing performance
- •Operating at SEA scale as a distinct advantage/learning environment
- 9:40 – 11:26
Joining a risky startup: evaluating value prop, market, and solvability
Crystal explains how she decided to join early-stage Gojek despite significant risk. Her decision hinged on a clear value proposition, market readiness, and confidence that operational obstacles (like driver smartphone adoption) were solvable.
- •“Little to lose” mindset early in career
- •Clear value prop: Indonesia traffic and motorcycle taxis
- •Supply existed but matching, pricing, and reliability were broken
- •Mobile adoption on the supply side was a risk, but fixable
- •Operational scale story: renting a stadium to onboard tens of thousands of drivers
- 11:26 – 13:01
Why super apps didn’t emerge in the U.S.
Crystal gives two key reasons super apps thrive in Southeast Asia but not the U.S.: different cultural trust in conglomerates and the practical constraints of mobile-first markets. Limited device storage and fewer alternative computing options push users toward one “good-enough” app.
- •Conglomerates in SEA are culturally normalized and trusted
- •U.S. consumers are more suspicious of consolidated data/power
- •SEA leapfrogged PCs—many households are phone-first
- •Storage/capacity constraints make app consolidation attractive
- •Super apps win by being “good enough” across many jobs-to-be-done
- 13:01 – 17:21
Scrappy growth: doing ‘crazy’ things and Wizard-of-Oz experiments
Crystal describes how Gojek validated ideas without building full features—using operational hacks, WhatsApp groups, vouchers, screenshots, and lightweight web flows. The core theme: quickly manifest the intended user experience, learn, then invest in scalable versions later.
- •“Do things that don’t scale” as a core growth habit
- •Wizard-of-Oz subscription test: drivers selling via WhatsApp + manual voucher fulfillment
- •Fake onboarding/UI tests via screenshot overlays and in-app messages
- •Using Typeform/web pages to test feature concepts without app releases
- •Build scalable primitives only after repeated scrappy validation
- 17:21 – 18:39
Using small data intentionally: experimenting even with ~30 users
Crystal argues early-stage teams should still run experiments with small samples. Small datasets often reveal the same directional truths; what improves with scale is precision, not the underlying trend.
- •Experiments are valuable even at sample sizes of ~30
- •Directionality vs. precision: scale refines confidence intervals
- •Early-stage is the cheapest time to test bold ideas
- •Focus on the step before the target outcome (leading indicators)
- •Design tests around concrete reasons someone converts, not just end metrics
- 18:39 – 22:21
Retention growth: stop treating retention as the goal—make it specific
Crystal clarifies her stance on retention: it matters, but it’s too vague unless translated into specific behaviors and barriers. She illustrates how improving retention often means fixing the moment right before conversion and building momentum through a better sequence of steps.
- •Retention is an outcome; teams need a more specific actionable target
- •Identify the “setup moment” before the “aha moment” (e.g., trust)
- •Example: using friend purchase signals to increase trust in new merchants
- •Use user psychology and momentum to carry users through high-friction steps
- •Retention gains often come from improving the last mile before conversion
- 22:21 – 25:01
What works vs. doesn’t in growth: copy, sure wins, and shortening painful paths
Crystal shares how teams waste time chasing unvalidated new features instead of leveraging clear bottlenecks. High-leverage growth work often starts with copy and messaging, then progresses to fixing the longest and most failure-prone paths users already attempt.
- •Red flag: building brand-new features with no evidence or tests
- •Copy is often the highest leverage when users bounce before first action
- •Look for “definite wins” where users try and fail—then fix the constraint
- •Shorten the longest time-to-aha paths among eventual converters
- •Example: speeding up/strengthening search at Kumu improved conversion substantially
- 25:01 – 27:55
Viability benchmarks: retention thresholds and the ‘friends & family’ bar
Crystal offers concrete rules of thumb for early viability using cohort retention. She distinguishes expectations for free vs. paid products and emphasizes that friends-and-family retention should be extremely high if the product truly solves a problem.
- •Free product: target ~60% week-1 retention, then flatten around that level
- •Paid product: ~20–30% retention benchmark depending on frequency
- •Very early stage: friends/family should be ~80%+ if it’s truly valuable
- •Cohorts matter: give a few periods for retention to stabilize/flatten
- •Cautionary tale: expansion can “pull forward” the small subset able to pay (e.g., credit card constraints)
- 27:55 – 31:58
Founders’ growth strategy: define ‘physics,’ identify levers, change one variable
Crystal explains her practical process for growth strategy: start with constraints (market, product, model, channels) and map how growth truly happens. Then optimize loops/funnels while avoiding overly complex bets that require many parameters to go right.
- •Step 1: articulate the ‘physics’—market, product, model, channel constraints
- •Include unconventional channels (e.g., physical presence and word-of-mouth)
- •Step 2: ensure any loop/funnel change fits those physics
- •Avoid changing too many variables at once; isolate impact
- •Real example: drivers as visible distribution + operational presence driving city expansion
- 31:58 – 36:38
Lever-based growth at scale: turning drivers into a GoPay acquisition engine
Crystal describes a targeted mechanism to grow GoPay by using drivers as incentivized salespeople at the exact moment of customer attention. A small backend check enabled tailored prompts and incentives, converting the operational network into a powerful distribution channel.
- •Map levers you already have but aren’t fully using
- •Build a small service to detect if a rider has used GoPay before
- •Real-time driver prompts: convert cash into wallet top-up with incentives
- •Leverage “captive attention” during the ride to explain value
- •Impact: driver-led GoPay acquisition became a major share of adoption
- 36:38 – 39:50
More growth unlocks: making new concepts familiar and reducing subscription churn
Crystal shares additional unlock patterns: make unfamiliar products legible via familiar metaphors, and fix churn by addressing the real cancellation reason. The AB InBev example demonstrates how a simple ‘pause’ option can outperform reactivation tactics.
- •Early unlock: packaging new concepts with familiar mental models (wallet as a “credit card”)
- •Tactical change can beat big feature bets when it targets the real barrier
- •Churn analysis: map cancellation reasons to reversible solutions
- •AB InBev D2C: adding “pause subscription” reduced permanent cancellation
- •Analogous patterns across industries (e.g., Airbnb listing snooze)
- 39:50 – 46:15
Why analytics efforts fail: instrumentation, context, and turning measurement into insight
Crystal argues analytics fails when teams track dashboards for status instead of designing data to answer ‘why.’ She distinguishes observations from insights and explains that good instrumentation captures context (properties) needed for segmentation and action.
- •Metrics should drive action; otherwise it’s ‘entertainment’ not ‘news’
- •Measurements/observations aren’t insights without context and causality
- •Insights come from hypotheses + segmented context + testing
- •Common failure mode: lots of events, few/no properties
- •Example: map load event should include supply, price, location, voucher, surge—enabling ‘why’ analyses
- 46:15 – 48:21
Doing analytics right: resources, recommended stacks, and practical tool choices
Crystal offers ways to learn event taxonomy and shares tool recommendations based on company stage and needs. She emphasizes mapping the user journey deeply and selecting tools that fit your maturity rather than overbuilding too early.
- •Learn by studying common flows (signup) and mapping ‘why would/wouldn’t’ a user act
- •Use existing guides (her post examples; Amplitude documentation)
- •Early dashboards: Google Data Studio; SQL-friendly option: Metabase
- •Mobile event + CRM needs: Clevertap; add Amplitude for deeper analytics as needed
- •Scaling data plumbing: Segment; experimentation: Eppo to accelerate decision-making
- 48:21 – 56:51
Building a growth team: what to hire for, how to structure, and interview approaches
Crystal explains how growth formed at Gojek as a “cleanup crew” filling gaps between core product work and real adoption constraints. She outlines when separate vs. embedded growth teams work best and how to hire growth talent with strong statistical and experimental rigor.
- •Early growth work often lives in operational/product gaps (OTP delivery, onboarding, protocols)
- •First growth hires should tackle known gaps—not come in to ‘invent’ growth from scratch
- •Separate growth team works best with very strong PMF and high velocity core product building
- •Hire for statistical intuition, sampling literacy, and bias awareness
- •Interview via take-home experiment design; prioritize quick, hacky, high-leverage execution
- 56:51 – 1:02:12
Generation Girl: expanding access for girls in STEM + how listeners can help
Crystal shares the origin and mission of Generation Girl, created in response to stereotypes and barriers women face in technical fields. She details programs, scale of impact, new initiatives for teachers, and specific ways people and companies can support.
- •Mission: empower girls (12–17) and students to explore STEM without bias
- •Programs: free classes, summer/winter clubs, weekly events, mentoring with engineers
- •Impact: thousands of students served; partnerships with major Indonesian tech companies
- •Kelas initiative: resources for teachers; part of MIT Solve; leverage teacher scale
- •Support: donations via GenerationGirl.org and in-kind support (licensed/enterprise software, iOS/dev tools)
- 1:02:12 – 1:03:09
Where to reach Crystal + final PSA on instrumentation
Crystal closes by sharing where to contact her and reiterating her core analytics message. She encourages teams to track journeys and experiences rather than just KPIs, so data can produce actionable insights.
- •Contact: CrystalWijaya.com (email available there)
- •Final call: instrument user journeys, not just KPI dashboards
- •Better properties/context lead to better insights and decisions
- •Actionable analytics creates better growth conversations
- •Episode wrap-up and goodbye
