Agency pre-send QA checklist for a new client

The first live sends for a new client are where agencies either earn trust or burn a domain. A pretty sequence means nothing if SPF is wrong, suppression is incomplete, or LinkedIn sits on the wrong identity.

Use this as a fail-closed gate. If a line fails, do not launch. Fix it, recheck, then send a small first segment.

GTM's motion assumes connected domains, send windows, limits, a compliance gate, and client-isolated workspaces. This checklist is the human layer before volume.

1. Domains and authentication

Confirm every sending domain the client will use:

• SPF, DKIM, and DMARC present and aligned for the From domain
• Tracking / link domains checked if you use them
• No surprise parked or lookalike domains in the rotation
• Warmup state matches the planned daily caps

Google's sender guidance expects authentication and, for bulk senders, DMARC plus workable unsubscribe behavior. Treat that as table stakes, not a nice-to-have.

Blacklist checks close to launch, not three weeks earlier. A clean check in week one can be stale by week three.

2. Compliance gate and required elements

Every commercial message needs accurate identity, a non-deceptive subject, a physical address where required, and a working opt-out path.

In GTM, the compliance gate is a database constraint: no send without a verdict and an unsubscribe URL. That is the right posture. Still verify with a test send:

• Unsubscribe link resolves and suppresses
• Headers match the client identity the prospect expects
• Sample body does not invent claims outside the client voice brief

FTC guidance is clear that companies remain responsible for messages sent by agencies acting on their behalf. Document who approved content before the first wave.

3. Suppression that actually blocks

Pull a current suppression master for the client and cross-check the launch audience immediately before dispatch:

• Unsubscribes and stop-reply contacts
• Hard bounces
• Complaints
• Open deals / current customers the client flagged
• Cross-client accidents if your stack ever shared pools (it should not)

Suppression that only lives in a spreadsheet is not suppression. It must survive purge and apply across campaigns. Re-enrollment after an opt-out is how agencies create complaint spikes.

4. Send windows and domain limits

Set windows to the recipient's working hours, not the agency's midnight hustle. Cap daily volume to the warmup ceiling per inbox. Ramp in writing: starting caps, weekly increases, and a kill switch if bounce or complaint rates move.

If email and LinkedIn both run for the client, define shared stop rules so a reply on one channel kills the other.

5. Identity isolation

Confirm the client workspace owns:

• Its email domains and mailboxes
• Its LinkedIn identities (credentials forwarded to the connection vendor, never persisted in a way that becomes a shared secret dump)
• Its ICP, entry rules, and voice brief
• Its suppression and provenance trail

Multi-client agencies fail here when "we just reused the agency mailbox." Isolation is not a UI preference. It is how you avoid sending Client A's pitch from Client B's domain.

6. Sample message review

Before first live volume, review a real sample set with a human who knows the client:

• Personalization grounded in prospect evidence, not fake research
• Claims present in the voice brief only
• CTA is a conversation, not a novel
• Sequence stops on reply (test it)
• No template that reads identical across ten prospects

Experiment Outbound's preflight framing is useful: automate authentication, hygiene, and scheduling; keep humans on claims and brand-risk judgment.

7. First-segment plan

Launch with a small, recent, verified segment. Define bounce and complaint thresholds that auto-pause. Name the on-call owner for the first 48 hours.

Record the launch assumption: list source, verification method, send caps, and who signed off. When something breaks later, you want a paper trail, not folklore.

One-page launch card

Copy this into the client workspace:

Any Block means no send.

Sources

• Cold email pre-launch checklist for agencies (OutboundQA)
• Pre-Send Cold Email Checklist (ForgeSend)
• Email Compliance Checklist (Adaptive Security)
• Preflight QA for AI Outbound (Experiment Outbound)

Building new client outbound with isolated workspaces and a compliance gate that refuses incomplete sends? Join the GTM waitlist.