Who You Learned From: Your Interviewee Log and Sample

• No names in reports. Quote people by role and code: “a junior who uses drop-in hours weekly (P06).” Use a name only if the person said yes to being named.
• Keep the key apart. If your team needs to know who P06 is (to follow up or send a thank-you), keep that name-to-code key in a separate, team-only file, not in the log. Decide as a team when you will delete it.
• Store it securely. One team-only shared sheet in your UT account. Not a personal phone note, not a public link, not a group chat.
• Collect only what you need. No contact details, ID numbers or birthdays in the log. Every extra field is one more thing to protect.
• Ask demographic questions respectfully. Ask at the end, explain why, and make it easy to skip: “To describe who we talked to, could I ask your year in school? You can skip this.”
• Be careful with sensitive topics. Health, income, religion, race or immigration status should not be in your log unless the challenge truly depends on it. Check with Lane or Jules first.
• Mind your AI use. Your course AI policy says no confidential client or personal data in public AI tools. That includes your log.

Need the consent opener and the basics of confidentiality? See Stakeholder Interviews, Consent and Ethics from BUAD 317.

Build your sample sentence in this order:

1. How many people and conversations, and when. “We interviewed 24 people in 27 conversations between September and November.”
2. By group. “13 students, 7 staff and 4 alumni.”
3. How you found them. “Most came through our client contact and our own outreach. 9 came by snowball referral from 4 starting points.”
4. Only the demographics that matter. “13 of 24 were women. 15 of the 24 were between 18 and 22.”
5. Who is missing. “We could not reach evening-only students.”

This is a rule of thumb for this course, not a statistics law. Your tool applies it for you.

Two more habits: describe patterns as “most of the students we talked to,” not “students prefer,” because you are not claiming to speak for every student. And skip significance tests. Qualitative research earns trust through rich detail and many voices, not p-values.

Most real projects mix both. That is fine, as long as your log shows it and your report says so.

What it is. At the end of a conversation, you ask the person who else you should talk to. Each new person can refer more, so the sample grows like a snowball. It works well for groups that are hard to find through official channels: night-shift workers, small business owners, people who stopped using a service.

How to ask for a referral:

• “Who else should we talk to who sees this differently than you do?”
• “Is there someone who has had a hard time with this, or who gave up on it?”
• “Would you be willing to pass our contact on to them, so they can decide whether to reach out?” Letting people choose protects their privacy.

Saturation, in plain terms. You are near saturation for a group when new conversations stop adding new themes: you can predict what the next person will say. Check it group by group, not for the whole sample. Ten students and one manager is saturation for nobody on the management side.

A simple test: after each conversation, write one line in the notes column on what was new. When the last two or three in a group add nothing new, say so in your methods. Do not claim saturation without that evidence.

Extremes and mainstreams. Your required text, The Field Guide to Human-Centered Design (IDEO.org), includes a method by this name. Talk with mainstream users, and also with people at the edges: heavy users, people who avoid the service, experts and first-timers. Extremes make needs easier to see, and mainstreams keep you grounded. Record in your log which is which, so your report can show the effort you made to include both.

Example (fictional): a campus career center’s drop-in hours.

“We interviewed 16 people in 17 conversations between September 2 and October 8: 8 students, 4 career center staff, 3 employer recruiters and 1 recent alum. Our client contact introduced us to staff and recruiters. We recruited students through outreach, at the drop-in line, and by snowball referral (6 students, all through one starting student). We shadowed the drop-in desk once and returned to one staff member to test a paper sign-up card. 8 of 16 participants were women. Limitations: our student voices lean toward one friend network, we heard from only one alum, and we could not reach evening-only students or recruiters from small companies.”

Notice what it does: counts, groups, sources, an honest snowball warning, and the gaps. The tool’s example produces the same warnings, so you can see where each sentence comes from.

• [ ] Every row has a code, not a name, and the name-to-code key (if any) lives in a separate team-only file.
• [ ] Consent is noted for every conversation.
• [ ] Every row has group, how found, method, date, length and interviewer.
• [ ] Snowball rows say who referred them, and no single starting point dominates.
• [ ] Every group you quote has at least 3 people, or you name it as a limitation.
• [ ] Optional demographic fields are few, relevant and allow “Prefer not to say.”
• [ ] Your sample sentence uses counts for small groups (“11 of 24”), not percentages alone.
• [ ] Who you could not reach is in your limitations.
• [ ] If you used AI, no names or confidential client data went into a public tool, and your AI statement says how you used it.