CPO Listening Tour Interview Guide

Nine questions, run the same way in every interview, so the answers are comparable across the org. The point is a ledger of falsifiable claims, not a word cloud. Pair this with the trust ledger and claim ledger (separate artifact).

Logistics

• 45 minutes, not 30. Thirty minutes produces the rehearsed answers. The real ones start around minute 25.
• One interview at a time. No group sessions. Group answers converge on the safest opinion in the room.
• You take the notes. Do not delegate. The act of writing is the act of noticing.
• Reserve the last 10 minutes for the kill question (Q9). Tell them up front it is coming so they can think.
• Same nine questions for everyone, with the per-function variants below swapped in where marked.
• No conclusions shared. You are allowed observations and questions. The moment you broadcast a thesis, your information supply gets curated to match it.
• Within 5 minutes of the interview ending, fill the capture form (bottom of this file). Same day or the signal is gone.

The nine questions

1. "Walk me through the last week of your actual work."

Why: The gap between the role on paper and the role in practice is your first map of the org. You want the calendar, not the job description. Listen for: How much of the week is meetings about work versus work. Who they had to chase. What they did that nobody asked for.

2. "What is the product actually good at, in your words?"

Why: You will get the pitch deck answer first. Wait. The second answer is the real one. Listen for: Whether they describe a customer outcome or a feature list. Whether their answer matches anyone else's.

3. "Tell me about the last decision that frustrated you. How did it actually get made?"

Why: This feeds the decision audit. You are mapping the real operating model, which is never the one on the wiki. Listen for: Who decided, what evidence was in the room, how long it took, whether the decision stuck.

4. "What number do you look at first in the morning, and do you trust it?"

Why: Tells you what the org actually optimizes and where the instrument gaps are. Listen for: Hesitation on "do you trust it." Hesitation is data. Also note when the answer is "I don't look at a number."

5. "Where does the product lose money or quality and nobody owns it?"

Why: Feeds the money audit and the quality audit. People three levels down usually know exactly where the leak is. Listen for: Named workflows. If they name one, write it down verbatim and ask who would own the fix if anyone did.

6. "What did the last person in my seat get right, and what did they avoid?"

Why: The avoided list is your inherited backlog of hard calls. The "got right" list tells you what to protect. Listen for: Patterns across interviews. One person's grievance is noise. Five people naming the same avoided decision is your month-three agenda.

7. "If you had my job for a quarter, what is the one thing you would change?"

Why: Surfaces who has been waiting years for someone to fix the thing they care about. These people become allies fast if you take them seriously. Listen for: Specificity. A specific answer means they have thought about it. A vague answer means they have given up or never cared.

8. "Who should I talk to that I probably won't think of?"

Why: The org chart routes you to the loud people. This question routes you to the load-bearing ones. Listen for: Names that come up more than twice across interviews. Interview those people next.

9. The kill question (last 10 minutes): "What should we stop doing? What is running on inertia?"

Why: The kill list at day 75 is built from these answers. Asking everyone the same kill question gives you frequency data, and frequency is your cover when you kill the thing. Listen for: Whether they name something they own. Someone willing to kill their own project is someone who runs on evidence. Note that separately.

Per-function variants

Swap these in for Q4 and Q5. Keep the other seven identical.

Engineering

• Q4 variant: "Which eval or quality number do you check before you ship, and who else looks at it?"
• Q5 variant: "Where is the codebase or the AI stack quietly expensive, in compute or in maintenance, and who owns that cost?"
• Extra listen-for: tech debt described as a moral failing versus a tradeoff. The first is exhaustion, the second is health.

Sales

• Q4 variant: "What number do you have to hit, and what does the product do that makes it harder?"
• Q5 variant: "What do you promise customers that the product cannot actually do yet?"
• Extra listen-for: the promise list. It is your shadow roadmap and your trust-repair list with customers.

Customer Success

• Q4 variant: "Which accounts are you worried about, and what is the earliest signal you saw?"
• Q5 variant: "What do customers work around manually that we pretend the product handles?"
• Extra listen-for: workaround stories. Every workaround is a discovery interview the org already ran and ignored.

Finance

• Q4 variant: "Can you tell me cost per outcome for our top three workflows, or only the aggregate cloud bill?"
• Q5 variant: "Where does product spend show up in the P&L in a way that worries you?"
• Extra listen-for: whether margin by workflow exists as a number anywhere. If it does not, you found your first instrument to build.

Board members

• Run as 30 minutes, not 45. Three questions only: Q2, Q6, and "What would make you nervous at my day-90 readout?"
• Why the nervous question: It tells you the objection to pre-wire before the readout, three months early.
• Extra listen-for: which metric each board member personally anchors on. Write it down per person. It shapes the readout scoreboard.

The post-interview capture form (5 minutes, same day)

Fill one per interview. These rows become the claim ledger.

# Capture: [Name], [Function], [Date]

## Claims they made (each as a falsifiable statement)
1. [Claim, in a form data could contradict. "Activation drops at the pricing page" not "onboarding is bad"]
2. [Claim]
3. [Claim]

## What they want from me
[One sentence. Everyone wants something. Knowing it is not cynicism, it is calibration.]

## What they fear
[One sentence. Often more honest than what they want.]

## Kill answers (Q9, verbatim)
- [What they said we should stop doing]

## Commitments I made (copy these to the trust ledger NOW)
- [Anything you said you would do, look into, or revisit. With a date.]

## Names they routed me to (Q8)
- [Name, why]

## One surprise
[The thing that did not fit your priors. If nothing surprised you, you ran the interview wrong.]

Frequency rule

After every five interviews, scan the capture forms for claims and kill answers that appear three or more times. Those go to the claim ledger marked high-frequency. By interview 20 you will have the skeleton of the day-90 readout without writing a word of it yet.