Lecture 16: How to Run a User Interview, with Emmett Shear

What this lecture covers

Emmett Shear, co-founder and former CEO of Twitch, teaches practical user-interview technique as the lecture's "talking to users" installment in the How to Start a Startup series. He grounds the talk in his own history: building Kiko without talking to any calendar users, building Justin.tv as a product for himself, and eventually running the user research that shaped Twitch's gaming pivot.

After watching, you should be able to choose which group of people to interview for a given idea, run an interview that avoids asking directly about features, and separate the goals people are trying to reach from the specific features they ask for.

Key ideas

• Who you talk to matters more than what you ask: interviewing broadcasters versus viewers of Twitch produced completely different, and differently useful, feedback.
• Don't build only for yourself: Justin.tv worked as a product for its founders' own reality show but stalled because they lacked insight into what a broader audience needed.
• Avoid asking about features directly: asking "would you use this feature" gets an artificially positive answer (the "horseless carriage effect"); instead, learn about current behavior and unmet goals.
• Interview at least six to eight people, and across different extremes of your target audience, before concluding whether an idea addresses a real problem.
• Talk to three groups: existing users of your product, users of competing products, and non-users, since each reveals different information, and non-users often reveal what blocks market growth.
• Validate cheaply before building: hack together a minimal version (such as a browser extension) rather than fully building an idea, and use willingness to pay as the strongest validation signal.
• Record interviews: recordings let you review conversations accurately and persuade the rest of the company by playing back a user's own words, which Shear says is more convincing than summarizing secondhand.
• Focus on underlying goals, not requested features: Twitch broadcasters asked for specific chat features, but the real drivers were money, stability, and reach, so Twitch invested in subscriptions and broadcast quality instead.

Walkthrough

Learning the hard way: Kiko and Justin.tv (0:00)

Shear recounts his first startup, Kiko, which the founders built without ever talking to a calendar user, and his second, Justin.tv, which worked because the founders were their own customer but stalled once they wanted to grow beyond that single use case.

Choosing who to talk to (4:44)

Shear explains how Twitch determined that broadcasters, not viewers, were the most important group to interview, because viewers followed the content and the broadcaster came along with it. He stresses that there is no fixed recipe for identifying your target user; it takes judgment.

Live exercise: picking interview targets for a note-taking app (5:50)

Using an audience-submitted idea for a lecture-focused note-taking app, Shear has the class practice identifying who they would interview, drawing out that students, IT administrators, and parents could each be relevant depending on who actually pays for the product.

Running a live user interview (12:03)

Shear interviews a volunteer student about her current note-taking habits (a mix of pen and paper, Google Docs, and Evernote) to demonstrate interview technique: asking about existing behavior and collaboration patterns rather than pitching a feature or product idea.

The trap of asking about features directly (16:25)

Shear explains why asking users to evaluate a proposed feature produces unreliable, overly positive answers, and why the goal of early interviews is to assess whether a real, sizable problem exists before deciding what to build.

Validating an idea cheaply (22:44)

He describes two validation paths: build a minimal, hacked-together version to observe real usage, or get people to commit money upfront, which he calls one of the strongest signals of genuine interest.

What Twitch learned from three groups of users (25:43)

Shear walks through condensed feedback from existing Justin.tv broadcasters, competitors' broadcasters, and non-broadcasters. He shows how the specific feature requests differed by group, but the underlying goals (income, stability, reach) were consistent, and those goals became Twitch's actual product priorities.

Audience questions on getting buy-in and reaching international or hard-to-reach users (34:49)

In the Q&A, Shear covers common interview mistakes (showing users your product, talking only to easily reached people), the persuasive power of recorded interviews for internal buy-in, the difficulty of researching non-English-speaking markets, and how the set of people worth interviewing shifts as a company grows, citing Twitch's later relationships with game publishers.

Before you watch

• No specific earlier lecture is required, though pairing this with the course's product-focused lectures gives useful context for why user research matters early in a startup's life.

Check your understanding

1. Why did interviewing broadcasters rather than viewers change Twitch's product direction?
2. What is the "horseless carriage effect," and why does Shear warn against asking users to evaluate specific features?
3. What are the two cheap ways Shear suggests for validating an idea before fully building it?
4. Why did Twitch prioritize investments like subscriptions and video stability over the specific chat features broadcasters requested?
5. Why does Shear say the group of users worth talking to changes as a startup grows?