I've been on the other side of this conversation a lot — a founder who has already been burnt once, trying to work out whether I'm different, and not really knowing what to ask.
The problem is that every agency's pitch looks the same. Everyone has a dashboard screenshot. Everyone has a reply rate. Screenshots are trivial to cherry-pick, so they tell you almost nothing.
Here's what I'd ask instead.
1. What's your worst campaign, and why did it fail?
The single most useful question. Anyone who has run outbound at volume has killed campaigns. If they can't name one, they either haven't run many or they aren't being straight with you.
What you want is a specific diagnosis — the list was too broad, the offer didn't match the segment, the domains were burnt before we started. Vagueness here is the tell.
2. How many domains and inboxes will I be sending from?
There's no single correct number, but there should be a considered answer tied to your target volume. "We'll figure it out" means they haven't done the arithmetic, and sending volume that outruns your inbox count is the most common way this work dies in week three.
3. Who writes the copy, and do they write per segment?
Templated copy sent to everyone is the fastest way to burn a domain. Ask to see two sequences written for two different segments of the same campaign. If they look like the same email with the industry swapped, that's your answer.
4. What does your warmup schedule look like?
There should be one, it should take weeks, and they should be able to tell you roughly when volume starts ramping. An agency that can start sending at full volume on day three is an agency that is about to burn your domains.
5. How do you verify the list?
Bounce rate is what gets you flagged. Ask what verification they run, what bounce rate they consider acceptable, and what they do with the contacts that come back risky rather than invalid.
6. What's your definition of a positive reply?
This is where most reported numbers fall apart. "Reply rate" includes "unsubscribe me." Ask specifically how many replies were actually interested, and how that's counted.
For reference, on a recent campaign of mine: 202 replies, of which 107 were positive. That ratio — roughly half — is the number worth comparing, not the headline reply rate.
7. What happens when replies start coming in?
Someone has to read them, sort them, and answer quickly. Ask whether that's you or them, and if it's them, what the response time looks like. A meeting booked four days after the reply is usually a meeting that doesn't happen.
8. What do I own at the end?
The domains, the inboxes, the lists, the copy, the sequencer account. If the engine lives entirely inside the agency's tooling, you are renting, and leaving means starting over. That may be an acceptable trade, but you should know you're making it.
9. What would make you tell me not to do this?
Outbound doesn't suit every business. If your deal size is small, or your buyer isn't reachable by email, or you have no case studies yet, cold email is a hard channel and someone honest will say so.
An agency that thinks outbound is right for literally everyone is selling, not advising.
The pattern underneath all nine
Every one of these is really the same question: has this person watched enough campaigns go wrong to recognise a problem before it costs you six weeks?
That's the whole job. The tooling is learnable in a month. Reading a campaign that's quietly failing — and knowing which of the four usual causes it is — takes volume and time. I've sent north of 5.5 million cold emails across 7 countries, and the thing that experience actually bought me wasn't better copy. It was pattern recognition on failure.
Ask the nine questions. The confident, specific, slightly unflattering answers are the ones you want.