Every new CPO gets handed the same advice: spend your first month on a listening tour. Forty coffee chats, open questions, lots of nodding. I have run that tour. It produces a word cloud. At day 30 you have a stack of notes, a list of themes the org already knew about itself, and no way to tell which are true. The listening is not the problem. The format is. So I stopped running tours and started running a ledger.
The nine questions, asked the same way every time
I ask the same nine questions in every interview, in the same order, regardless of who is across the table. The repetition is the point. When forty people answer identical questions, divergence becomes measurable. The set includes "What should I kill?" (four words, and people light up because nobody has asked them in years), "Where does this org lie to itself?" (phrased exactly like that, because the flinch tells you as much as the answer), "What did my predecessor get right that everyone now dismisses?", and "What is the eval score on our most important AI feature, and which way is it trending?" as the maturity X-ray. What is deliberately missing is "what are our biggest opportunities" and every other question that invites a prepared narrative. You will get the prepared narrative anyway. Do not spend a question on it.
The ledger schema
Every claim from every interview gets a row with five fields. The Claim is one falsifiable sentence: "Enterprise churn is driven by onboarding time," not "customers are frustrated." The Source is who said it, their function, and the date, because anonymous claims cannot be weighed. Confidence starts low and only evidence moves it, never repetition, so three people repeating one hallway theory is one claim, not three. Contradicts holds the row numbers this claim conflicts with, and it is the most important field in the file. Evidence needed names what would settle it, and if you cannot fill that field, the row is an opinion.
Contradictions are the product
By interview fifteen, the contradictions field starts to glow. Sales says the big account churned because engineering shipped late. Engineering says the deal was sold on features that never existed. Customer success says the account was never onboarded properly and both other stories are downstream of that. Three functions, one event, three causal stories. That gap maps the real operating model: where information dies, who controls which narrative, which version of reality the CEO has been buying. Agreement across functions is usually just the official story echoing. Disagreement is information. Keep a separate contradiction list sorted by stakes, and resolve the top three publicly in month two.
If you are in your first month, take your existing interview notes and convert the ten strongest assertions into ledger rows. The conversion itself shows you how much of what you collected is checkable and how much is vibe.