# Four Questions Before You Analyze

> From *The Decision Product* (Chapter 5). One page. Fifteen minutes.
>
> You need four answers before you spend serious time. If you cannot get them
> in a fifteen-minute conversation, that itself is information. It means the
> choice is still unspecified, or that nobody who can specify it will talk to
> you. Either way, opening the warehouse is a stall.

---

### 1. Who will use this to change an action?

Not who is *interested*. Who can commit, or who will run a rule.

Interest is cheap. Authority is not. If the answer is "the S&OP meeting," you
do not have an owner. You have an audience.

**Answer:** ____________________

---

### 2. What are they choosing among, including the status quo?

If they cannot name two commitments of the same kind, you are not yet on a
decision.

"Do a study or do nothing" is not two commitments of the same kind.
"Keep the current rule or change the rule" is.

**Answer:** ____________________

---

### 3. When does the choice go stale?

The meeting time is usually the wrong date. The freeze, the contract window,
the promotion start, the quarter-end cash event: those go stale.

**Answer:** ____________________

---

### 4. What would have to be true for them to pick differently?

This tells you what evidence is worth gathering. It is also the question most
likely to reveal that the decision was already made before you were asked.

**Answer:** ____________________

---

## When not to analyze at all

Three cases where reframing is not enough and the right move is to decline:

**The decision is already made and the analysis is wanted as cover.**
Say that you will document the rationale, rather than produce a study that
pretends to be an inquiry.

**Nobody in the conversation can make the choice.**
The deliverable is a short note identifying who can, not six weeks of work
addressed to nobody.

**The request is for a demonstration rather than a decision.**
"Build an agent" is the current form of this. The interesting questions are
whose work changes, what the system may do without a human, and what a failure
costs. Until those are named, a demonstration is a demo. Demos can be useful.
They should not consume a freeze date.

---

The Decision Product  ·  thedecisionproduct.com