Lecture 4 - Building Product, Talking to Users, and Growing (Adora Cheung)

Cheung advises blocking long, uninterrupted stretches (e.g., full days) rather than scattered hours, because context switching kills momentum in building and learning loops.

You should articulate the problem in one sentence and confirm it’s real for others by talking to people; her Pathjoy example shows how a ‘big mission’ can still map to a product you wouldn’t personally use.

Short-term immersion—getting a job, doing the service, reading everything (competitors, filings, earnings calls)—reveals the real inefficiencies and earns user trust in your competence.

Even if the endgame is ‘everyone,’ early success comes from cornering a specific segment you can serve extremely well, then expanding once you have product-market clarity.

Map discovery → landing → purchase/usage → post-use evaluation/review so your initial build solves the full experience that determines retention and referrals.

Shipping a bare feature without a complete, satisfying experience leads to churn; Homejoy’s traction improved when they simplified positioning to a functional promise (“get your place cleaned for $20 an hour”).

From #showhn to local parent mailing lists to literally intercepting customers at street fairs, the goal is to go from 0→1→10 users fast so you can start the feedback/iteration engine.

Surveys skew to extremes; in-person conversations (coffee/drinks) yield nuance, and paid users are often more honest than friends/family because money raises feedback stakes.

Track retention cohorts and CLV over time; if you can’t wait for full retention data, use reviews/ratings or NPS as faster signals of whether iterations improve stickiness.

Manual onboarding/interviewing taught Homejoy which questions predicted cleaner quality, enabling later automation; automating too early locks you into bad process and slows iteration.

Cheung prefers “temporary brokenness” over “permanent paralysis,” and warns against the “Frankenstein approach” of implementing every requested feature instead of diagnosing the underlying need.

Focus on one channel for a full week, iterate continuously, revisit formerly-failed channels as your economics change, and for paid growth ensure CLV exceeds CAC with a safe payback window (often ~3 months).

If you see multiple weeks of flat/negative growth while working hard, something is fundamentally wrong; use an expected growth plan and pivot when users won’t stick or the unit economics don’t work.

Switching is hard when users are ‘comfortable’; Homejoy converted users during urgent needs (next-day availability) and then let accumulated convenience advantages (online payment, easy rescheduling) do the rest.