Email deliverability
Deliverability is whether a message reaches the inbox rather than the spam folder or nowhere. The ranking that matters is unpopular: who you send to and how they react outweighs everything else, authentication is necessary but buys nothing on its own, and the subject line most people worry about first is near the bottom.
The ranking#
- Who is on the list, and how they got there. Recipients who asked for the mail open it, reply to it and do not complain. No configuration compensates for a list that was bought, scraped or guessed.
- Complaint rate. A fraction of a percent is the scale that matters. Complaints are weighted heavily because they are unambiguous — the person pressed a button that says so.
- Engagement over time at that provider. Large mailbox providers assess a sender against their own users' behaviour, which is why placement can differ sharply between two providers on the same day.
- Authentication. SPF, DKIM and aligned DMARC. Failing these will hurt you; passing them is a floor, not an advantage.
- List hygiene. Hard bounces removed promptly, and spam traps avoided — both the recycled kind (a real address a provider reactivated as a trap) and the pristine kind (never a real person, so its presence proves the address was not collected from a human).
- Identity consistency and volume shape. A stable sending domain, marketing separated from service mail, and no sudden order-of-magnitude spikes from a domain with no history.
- Infrastructure hygiene. Valid reverse DNS, TLS on the connection, a HELO name that resolves.
- Content. Link domains that match the sender, no URL shorteners hiding the destination, text as well as images. Real, and last.
Why measurement is harder than it looks#
Open rate has been an unreliable signal since mailbox providers began proxying and prefetching remote images: an 'open' may be a machine, and a real reader may never register one. Any deliverability judgement resting on opens alone is resting on a number whose meaning changed under it.
Connect's own reader makes the same choice deliberately, in the other direction. Incoming mail is sanitised aggressively and remote images are proxied rather than fetched by the reader's browser, so opening a message in Connect does not tell the sender you opened it. If you are the sender, expect the same treatment from everyone else.
What remains measurable: bounce classes, complaint reports where a feedback loop exists, replies, and the aggregate authentication results in DMARC reports. Replies are the strongest signal of all and the one nobody can fake on your behalf.
The failure that looks like success#
- Bounce rate near zero, reply rate near zero
- Mail is being accepted and filtered. There is no bounce to count because nothing was rejected.
- Placement good at one provider, poor at another
- Normal, and it points at engagement rather than configuration — the same authentication is passing at both.
- A sudden drop after a good month
- Usually a volume change, a new sending source, or a batch from a different acquisition path mixed into an established list.
- Everything fine except one recipient's employer
- A corporate filter, tuned locally. Provider-level reputation work will not move it; a conversation with that recipient's administrator might.
Does Connect use deliverability controls?#
Connect does not run a sending platform, does not hold IP reputation and does not send bulk campaigns. Mail leaves through the workspace's own mailbox, so the domain's reputation is the domain's, unchanged by connecting an assistant to it.
Two product decisions bear directly on the top of the ranking. First, Connect does not guess email addresses: no pattern-generated address is produced, tested or contacted, and a prospect with no discoverable address is kept as a researched prospect with a different next action. That deliberately costs volume, and it is the single most effective deliverability measure in the product, because guessed addresses are how senders find pristine spam traps and accumulate hard bounces.
Second, volume is bounded rather than unlimited: a daily allowance governs how much mail is processed, and when it is spent, work is held rather than dropped, with the refusal shown in Needs You. A held queue is visible; a silent spike is not.
Everything else on the ranking — the DNS records, the domain strategy, the decision to separate marketing from service mail — sits with the domain owner. Connect does not configure any of it and does not pretend to.
Questions#
We authenticated everything and placement did not improve. Why?
Because authentication is a floor. It proves the mail is from who it claims to be from, which stops one class of rejection and answers none of the questions a filter asks about whether recipients want it. If placement did not move, the constraint is above authentication in the ranking — almost always the list or the complaint rate.
Does a dedicated IP help?
Only at volume, and it can hurt below it. A dedicated address with little traffic gives providers too little evidence to form a view, and any single bad batch is unmixed with anything else. Below a substantial and regular volume, shared infrastructure with a good operator is usually better.
Should marketing and service mail come from the same domain?
Separating them — commonly by subdomain — limits the damage a marketing reputation problem can do to password resets and invoices. It also lets you hold different suppression semantics on each stream, which is the practical reason to do it even at modest volume.