Outreach email versus a customer reply
A reply answers somebody who wrote to you. Outreach starts a conversation with somebody who did not. They share one send boundary and nothing else: outreach must additionally clear provenance, evidence freshness, a role address, an owner-reviewed offering, domain and daily caps, suppression, opt-out state, sender identity and a reputation state. A reply clears none of those.
Why they are two paths#
The difference is consent, and everything else follows from it. Somebody who emailed a business has invited an answer; the risk in replying is getting the content wrong. Somebody who has never heard of the business has invited nothing; the risk in writing to them includes whether the message should exist at all. One gate cannot express both without being either too loose for the first case or unusable for the second.
So the two are separate in the code and stay separate. Merging them is the change that looks like tidying up and is not: the moment a reply runs through the outreach gate, ordinary customer mail starts being refused for want of a research record, and the moment outreach runs through the reply gate, cold mail goes out with none of the checks that make it lawful.
What each path must clear#
| Check | Customer reply | First outreach |
|---|---|---|
| Do-not-contact on the recipient | Yes | Yes |
| Mechanical claim check — no invented price, service level or warranty | Yes | Yes |
| Channel autonomy mode, scopes and the workspace's own gates | Yes | Yes |
| The daily sending allowance and the shared send boundary | Yes | Yes |
| Verified provenance for the address, and evidence no older than the configured age | No | Yes |
| A role or business address, unless a named work address was authorised on upload | No | Yes |
| A qualification score above the campaign's floor | No | Yes |
| An active, owner-reviewed offering to write about | No | Yes |
| Deep research that is fresh and free of hard disqualifiers | No | Yes |
| A cap on first contacts to one recipient domain in a day, and on new prospects a day | No | Yes |
| Provider suppression state, and promotional opt-out state | No | Yes |
| A verified dedicated sending identity, and the campaign's chosen mailbox | No | Yes |
| A reputation state that is not red, and not yellow where the workspace holds on yellow | No | Yes |
The list is the policy. Shortening it in the code would not shorten the policy; it would only move the decision somewhere nobody can audit.
Hard failures and live-only failures#
Not every unmet condition means the same thing. A failure about content, provenance or do-not-contact is hard: the message must not exist in that form. A failure about the sending infrastructure — an unverified dedicated identity, a suppression state that could not be read, a yellow reputation, a mailbox that is not the campaign's — blocks a live send while still allowing the research and the draft to be reviewed.
That split is what makes the gate usable. A business setting up outreach can see exactly what a first message would say before its sending identity is finished, and cannot accidentally send it.
What outreach mail is allowed to look like#
- Plain text, within a fixed length. Nothing enclosed.
- No tracking pixel, and no invented
Re:orFwd:prefix to imply a conversation that never happened. - Exactly one recipient, so the unsubscribe resolves to a real person.
- A working way out in the first message, carried by the provider's list management — see Unsubscribe.
- Nothing commercial that the workspace's Knowledge does not support, which is the same rule a reply obeys.
When the prospect answers, their reply is an ordinary inbound message: it is triaged, grounded and drafted like any other, and it stops the sequence. From that point the relationship is a conversation rather than a campaign, which is the transition described in Prospecting in Connect.
Questions#
Can I skip the outreach gate for one prospect I already know?
If there is an existing relationship, the message is a reply rather than outreach and takes that path. What is not available is running cold mail through the reply path because the gate is inconvenient.
Why does outreach refuse a named person's address?
A published role address and a named individual's address are different in law in much of the world. A named work address is allowed where it was authorised on upload; otherwise the safer default holds.
Does a customer reply ever need research?
No. It needs grounding — approved Knowledge, memory and the conversation — which is a different thing from the evidence a first approach requires.