What to ask, what the answers mean, and the seven patterns that predict a bad engagement before you sign.
TL;DR
Most of what separates a good outbound partner from a bad one is visible before you sign, if you know which questions produce answers that cannot be faked. This is the diligence list, drawn from running this work and from regularly inheriting accounts from providers that did it badly.
Scope varies more than pricing does, and the gaps are where engagements go wrong. Settle what is included before you compare numbers, because two proposals at the same monthly figure can differ by half the work.
| Usually included | Usually not, unless you ask |
|---|---|
| List building and enrichment | CRM administration and hygiene |
| Copy writing and iteration | Taking the meetings |
| Sending infrastructure and warmup | Proposal and pricing support |
| Campaign execution and monitoring | Nurture and long-term follow-up sequences |
| Reply routing and classification | Anything past the booked meeting |
| Deliverability management | Positioning and offer design |
This is the fastest disqualifier on the list. Apple Mail Privacy Protection pre-fetches images, security scanners click every link, and bot traffic inflates the rest. A partner leading a weekly report with open rate is either not paying attention or is counting on you not to.
Apple MPP, corporate security scanners, and link-checking bots. The metric was never reliable and stopped being directionally useful in 2021.
What should be in the report instead: replies, positive replies, meetings booked, bounce rate, and cost per meeting. Those are countable and hard to flatter.
Ask whose name the domains are registered in and whose account the mailboxes live in. There are two answers and they lead to very different places. If the infrastructure is yours, you leave with a warmed asset and the engagement is renewed on its merits. If it belongs to the provider, leaving costs you the entire warmup period, and both of you know it during every renewal conversation.
The related question is whether your domains are shared with other clients. Shared sending pools mean another company's bounce rate becomes your reputation problem, and you will never be told which company or when.
Every email going out under your name should be reviewed by you before it sends, at least until you have reason to stop reading them. A partner who resists this is protecting a process that does not survive inspection. The reasonable version of the objection is about speed, and the reasonable answer is an approval step that takes a day, not a waiver of the right to see what is being said in your name.
This is the request that separates claims from practice: ask to see five real emails sent to five different prospects in the last thirty days. Not case studies, not a deck, actual sends. If the five differ only in the company name and first name, personalization is a checkbox rather than a method, and the reply rate will tell the same story in about six weeks.
An order of magnitude, on identical infrastructure and identical lists. Copy is not the first thing to fix, but it is where the ceiling is.
The common objection is client confidentiality. It is answerable: redact names, domains, and any identifying detail, and show the structure. A partner unwilling to show redacted copy is telling you the structure is the problem.
Bounce rate is the cheapest quality signal in outbound and the one with the longest tail of consequences. Unverified sending damages your sender reputation, not the provider's, and reputation recovery takes weeks. Ask what the bounce threshold is and what happens when a campaign crosses it.
Above this, mailbox providers start treating the domain as a spam source, and the damage outlasts the campaign that caused it.
| Contract structure | What it signals |
|---|---|
| Month to month | Confidence, and a provider that expects to be judged monthly. Sometimes too short for outbound to prove anything. |
| Three months with performance benchmarks | The standard shape. Long enough for warmup and iteration, short enough to leave. |
| Six months, benchmarks defined | Reasonable for complex ICPs where ramp genuinely takes longer. |
| Twelve months, no performance clause | Protects the provider from being measured. The most common structure among the accounts we inherit. |
| Auto-renewing with a long notice period | Read the notice window carefully. Ninety-day notice on an annual term means deciding to leave before you have full-year data. |
Four to five weeks from kickoff is normal and mostly consists of warmup, which cannot be compressed. A promise of live campaigns inside two weeks means one of three things: warmup is being skipped, the domains are pre-warmed and shared, or the provider is describing setup rather than sending. All three are worth knowing before signing, and only the third is benign.
| Timeline | What should be happening |
|---|---|
| Week 1 | ICP definition, domain purchase, DNS configuration, mailbox creation |
| Weeks 2-3 | Warmup running, list building, copy drafted and reviewed with you |
| Weeks 4-5 | First sends at low volume, deliverability monitored, copy iterating |
| Weeks 6-8 | Volume ramping, first meetings, first real read on reply rate |
| Weeks 9-12 | Enough data per variant to draw conclusions and change something on purpose |
The last one is the most useful and the least expected. A partner with no answer either has not thought about where the channel fails or is unwilling to say so to someone holding a budget.
| Warning sign | What you want to hear instead |
|---|---|
| Open rates lead the report | Replies, positive replies, meetings, bounce rate, cost per meeting |
| Domains registered to the provider | Domains in your name, in your account, yours on exit |
| Specific meeting counts promised in month one | A ramp curve, with the warmup period named |
| Copy sends without your review | An approval step, with a turnaround you can live with |
| No stated bounce threshold | A hard number and a documented response when it is crossed |
| Five sample emails that read identically | Five that share a structure and nothing else |
| Twelve-month lock-in, no benchmarks | Three to six months with performance defined in writing |
We build and run these systems embedded with your team.