Agency pitches are optimized for the questions buyers usually ask. Everyone has a nice deck, a process diagram, and a case study. None of that separates the good from the plausible.
These six do, and every one is answerable on a first call.
The most useful question in the whole conversation, because the answer is verifiable later.
You want names and roles, not "our team". Find out who is on the call because they will run your account and who is on the call because they close deals. Both are legitimate, but you should know which is which before you sign, not after the third status meeting with someone you have never met.
Ask whether any of it goes to contractors, and if so, which parts. There is nothing wrong with a specialist bench. There is something wrong with finding out about it by accident.
Every engagement has a client-side cost: access, approvals, subject-matter time, someone to answer questions. An agency that has done this before can tell you roughly what that is. One that says "very little, we handle everything" is either not doing much or has not thought about it.
This question also predicts the most common way engagements fail, which is not agency incompetence. It is the client not having the internal time the work assumed.
A firm that does everything usually does a few things well and the rest passably, with a subcontractor filling the gap.
The answer tells you where the real expertise sits. It also tells you whether anyone there has made a strategic decision, which is a reasonable proxy for whether they will make good ones on your behalf.
Listen for whether an actual scenario comes back.
Every honest service business has a version of this: a client too early, a problem that is not theirs to solve, a fix that costs less than the engagement. If nobody can name one, either they have never turned work down or they are not going to tell you about it.
Not the twelve-month outcome. The ninety-day signal.
Real answers are specific and often unglamorous: a tracking gap closed, a data model in place, a reporting view that did not exist, a first cohort through a new funnel. Answers built entirely on lagging revenue metrics are a way of deferring accountability past the point where you would still cancel.
Ask this before you start, when it is a hypothetical and nobody is defensive.
You want to hear that accounts stay in your name, that documentation exists, and that you could hand the work to someone else. An agency structured so that leaving is expensive has made a choice about how it retains clients, and you should know which choice that is.
None of these are trick questions and a good agency will enjoy them. What you are measuring is whether the person answering has done the work or only sold it.
The tell is specificity. Someone who has run twenty of these engagements answers with a scenario. Someone who has not answers with a principle.