Needfinding

Lab setup

First, make sure you have completed the initial setup.

If you are part of a course

  1. Open Terminal. Run the update command to make sure you have the latest code.
    $ mwc update
  2. Move to this lab's directory.
    $ cd ~/Desktop/making_with_code/pilot/labs/lab_needfinding
    

If you are working on your own

  1. Move to your MWC directory.
    $ cd ~/Desktop/making_with_code
    
  2. Get a copy of this lab's materials.
    git clone https://git.makingwithcode.org/founders-studio/lab_needfinding.git

Most failed products don't fail because they were built badly. They fail because they solve a problem nobody really had. Designers avoid this by starting with needfinding: going out to understand people's needs by watching and listening to them in their own lives, rather than by asking what they want or guessing. In this lab, you will learn to do ethnographic needfinding, and find out what AI can (and can't) add to it.

By the end of this lab, you will be able to:

Why needfinding

๐Ÿ‘ Read examples.md, which describes three products that were built on assumptions nobody checked.

๐Ÿ’ป Discuss with your group: in each case, what did the designers assume? What would they have learned if they had watched people use the product (or live without it)?

Needs are different from solutions. "I need an app to remind me of deadlines" is a solution. The underlying need might be "I need to feel in control of my week," or "I need my teachers' deadlines in one place." People are usually better at telling you their problems than at designing solutions, and the most valuable needs are often ones people don't say out loud. You find them by noticing workarounds (sticky notes, spreadsheets, group chats) and contradictions between what people say and what they do.

Interviewing

A good needfinding interview is more like a conversation than a survey. Some guidelines:

๐Ÿ‘ Open interview_guide.md. It's a template with sections for introductions and consent, warm-up questions, story questions, and wrap-up.

๐Ÿ’ป Choose a topic with your team. You might start from the problem in your pitch from the Tooling lab, or from an area your team cares about. Draft an interview guide, by hand.

Using AI to prepare

๐Ÿ’ป Ask Claude Code to critique your interview guide using the guidelines above. Ask it to point out leading questions, yes/no questions, and hypotheticals. Revise the guide, but only take suggestions you agree with.

๐Ÿ’ป Practice your follow-up questions by asking the agent to role-play an interviewee. For example:

Role-play a high school junior who works a part-time job. I'm going to interview you about
how you manage your schedule. Give realistic, short answers, and don't volunteer much unless
I ask good follow-up questions.

Take turns being the interviewer. Your teammates should watch and note where the interviewer could have followed up.

Fieldwork

Now for the real thing. Each team interviews three to five people who are in your target group and are not classmates.

๐Ÿ’ป Conduct your interviews. Take notes in notes/, one file per interviewee (notes/p1.md, notes/p2.md, ...). If you record, transcribe the recording, and keep the transcript with your notes.

๐Ÿ’ป Right after each interview, while it's fresh, add a section to the notes called Surprises: anything you didn't expect.

Synthesis

Interview notes are full of both observations (what people said and did) and interpretations (what you think it means). Keeping them apart helps you notice when you are jumping to conclusions.

๐Ÿ’ป Go through your notes and mark each line as an observation (O) or an interpretation (I).

๐Ÿ‘ Open empathy_map.md. An empathy map has four quadrants: what people say, do, think, and feel. Says and does come straight from your observations; thinks and feels are interpretations, so each one should point to the observations that support it.

๐Ÿ’ป Fill in the empathy map as a team.

๐Ÿ’ป Ask the agent to read your notes and suggest themes: ideas that come up across several interviews. Then check every theme yourself. Does it really appear in the notes? Did the agent miss anything important, especially surprises?

๐Ÿ’ป Write insights.md: three to five key insights, each a short statement of something you learned about your users' needs, followed by the evidence (quotations and observations, with interviewee codes) that supports it.

Synthetic users

Some companies now sell "synthetic users": AI models that pretend to be your customers, so you can skip the interviews. Let's test the idea.

๐Ÿ’ป Start a fresh session and ask the agent to play five people in your target group. Interview them using your guide. Save the transcript in synthetic.md.

๐Ÿ’ป Compare the synthetic interviews with your real ones. Did the synthetic users mention any of your surprises? Were their stories specific? What did the real interviews include that the synthetic ones didn't?

Deliverables

๐Ÿ’ป Answer these questions in questions.md:

  1. What was the most surprising thing you learned in an interview? Why didn't you expect it?
  2. Describe a follow-up question that led to something important.
  3. How did AI help you in this lab? Where was it unhelpful or misleading?
  4. Would you trust a product designed using only synthetic users? Why or why not?