How to Start a Startup

HTSaS · Chapter 03

Talking to Users

The habit that never stops mattering: sit with users, watch what they do, and separate what they say from what they need.

After this lesson

  • Run user conversations that surface real behavior, not politeness
  • Build a weekly cadence of user contact that survives past launch
  • Turn raw conversations into a ranked, falsifiable pain list

The habit Chapters 1 and 2 depend on

Chapter 1 told you to find a specific problem; Chapter 2 told you to launch small. Both only work if you're actually talking to the people who have the problem. Lecture 4 (Adora Cheung) covers this alongside product and growth; Lecture 16, later in the original course, is Emmett Shear's dedicated masterclass on the interview mechanics themselves — pulled forward here because the habit needs to start now, not in week 16.

Talking to users never graduates out of the job

Founders who stay closest to users for the longest tend to win. This isn't a phase you finish after 'validation' — it's a founder habit for as long as you run the company, from first user to IPO.

Read the original

Lecture 4 — Building Product, Talking to Users, and Growing — Adora Cheung

Extract Cheung's framework for going from zero users to many — pay attention to how much of it is talking to (and personally serving) early customers.

This chapter is a working summary. When the idea matters, read the source once — then come back and do the practice.

Read the original

Lecture 16 — How to Run a User Interview — Emmett Shear

Pull Shear's specific interview questions and anti-patterns — this is the tactical companion to the 'signal or noise' decision below.

This chapter is a working summary. When the idea matters, read the source once — then come back and do the practice.

Watch behavior, not opinions

  1. Ask about specific past behavior ('walk me through the last time you hit this problem'), not hypotheticals ('would you use a tool that...')
  2. People are polite by default — enthusiasm without a follow-up action is not signal
  3. Watch someone actually use the product live whenever you can; screen-share or sit beside them
  4. Notice what they try to do that the product doesn't support — that's often the real job

A conversation that produces signal

  1. Open broad: how do you currently handle [the general area]?
  2. Narrow to the last specific instance, with details and emotion
  3. Ask what they've tried before and why it fell short
  4. Ask what they'd pay, or what budget exists, before you pitch anything
  5. Only pitch your solution at the very end, if at all — and watch their reaction, don't sell

Signal or noise?

After a batch of user calls, what's the pattern?

Case

Adora Cheung cleaning houses herself

Before Homejoy hired a single cleaner, Adora Cheung and her co-founder personally cleaned customers' homes to fulfill the company's earliest orders.

Takeaway: She wasn't just testing demand — she was standing inside the exact moment a customer's problem occurred, catching objections and friction points a survey or support ticket would never surface. Doing the job yourself first is a talking-to-users technique, not a separate step.

Turn conversations into a decision

  1. Log every conversation the same day — memory decays fast and gets rationalized
  2. Track what changed in your hypothesis after each call, not just what was said
  3. Re-rank your top pain list after every 5–10 conversations
  4. Decide explicitly: persevere, sharpen the persona, or pivot the problem

Practice