API docsAgent tools

Deliverability mechanics

Email deliverability

Learn Reviewed · not yet fact-checked Checked against sources 2026-08-03

View as Markdown

Mailbox providers score every incoming message on sender reputation, authentication, content, and recipient engagement, then route it to the inbox, the spam folder, or rejection. Delivery and inbox placement are different things — a message can be accepted and still never be seen. Senders control authentication, volume, list quality, and engagement.

What it is

Deliverability is the probability that a sent email reaches the recipient's inbox rather than the spam folder or silent rejection. It is decided per message, per recipient, by the receiving mailbox provider — not by the sender's tooling.

Why it works this way

Mailbox providers optimize for their users, not for senders. Every incoming message is scored on layered signals:

  1. Connection and infrastructure — the sending IP's history, reverse DNS, TLS.
  2. Authentication — does the message pass SPF, DKIM, and DMARC for the claimed domain?
  3. Sender reputation — the provider's running score for the domain and mailbox, built from past behavior. See Sender reputation.
  4. Content — links, attachments, HTML weight, spam-pattern phrases.
  5. Recipient engagement — do this provider's users open, reply to, and keep messages from this sender, or delete and report them?

The scores combine into a routing decision: accept to inbox, accept to spam, defer, or reject. Providers deliberately keep the exact weights secret and shift them over time.

How it behaves in practice

  • Delivery rate ≠ inbox placement. "Delivered" only means the receiving server accepted the message. Placement — inbox vs spam — is invisible in standard delivery metrics and must be measured with seed tests or engagement proxies.
  • Reputation is per domain, per mailbox, and per IP — damage to one sending mailbox can spread to its domain.
  • Engagement compounds. Replies are the strongest positive signal a cold sender can earn; spam reports are the strongest negative one.
  • Volume spikes read as abuse. A mailbox that jumps from 20 to 500 daily sends looks like a compromised account. This is why warm-up exists.
  • Thresholds appear at provider policy level: bulk-sender rules (such as Gmail and Yahoo's 2024 requirements) mandate authentication, one-click unsubscribe, and spam-rate ceilings around 0.3%.

Common misconceptions

BeliefReality
"Delivered means it was seen"Delivered means accepted; it can still land in spam
"Spam words are the main filter"Content matters less than reputation and engagement
"One good IP fixes everything"Domain reputation follows you across IPs
"Deliverability is set once"It is a running score that decays and recovers with behavior

How Reply implements this

Reply's deliverability capability manages the sender-controlled parts:

  • Volume caps. One email account sends at most 400 emails per day on paid plans (200 on trial), inside schedules and the sending-schedule rules.
  • Ramp-up by default. Ramp-up mode is active for every newly added email account; the default curve starts at 10 emails per day and adds 3 daily until it reaches 100.
  • Human-like pacing. A configurable delay spaces out individual sends; the default account safety setting sends each email with a random delay of 300-450 seconds.
  • Warm-up via MailToaster. Peer-network warm-up runs through MailToaster, included with paid subscriptions. See email warm-up.
  • Monitoring. A Domain Setup health checker validates the domain's DNS records, and the Spam Monitor tab reads your domain's Gmail spam rate from Google Postmaster Tools — the number to hold below the 0.3% ceiling. Reply's recommended deliverability check via the Mail Tester service is scored from 0 (bad) to 10 (excellent), and Reply recommends custom tracking domains over the default shared ones.
  • Automatic stops. Email steps pause for the rest of the day when a sequence's bounce rate crosses 8% (calculated per 24 hours, once at least 200 emails have been sent) — see bounce handling — and replies halt follow-ups per the sequence's reply-handling setting.

Mailbox selection documents how sends are distributed across accounts.

Where this breaks down

No sender-side tool can override a receiving provider's decision. If a domain's reputation is burned, the only fixes are time, reduced volume, and better targeting — or a new domain warmed from zero. Placement testing is also inherently sampled: seed inboxes approximate, but do not equal, your real recipients.

Build with Reply