Email Deliverability Metrics: What to Track, Calculate, and Improve

Email deliverability metrics tell you whether your messages merely reached a receiving mail server or actually earned a place where subscribers can see and act on them. The difference matters: a strong delivery rate can coexist with poor inbox placement, rising spam complaints, and declining email revenue.

This guide explains the email deliverability metrics worth tracking, the formulas behind them, what each metric can and cannot prove, and the actions to take when the numbers move in the wrong direction. It applies to transactional email, newsletters, lifecycle campaigns, and product notifications, though the right thresholds and segmentation differ by mail type.

Start with the distinction: delivery is not deliverability

The most common reporting mistake is treating delivery rate as proof of deliverability. They are related, but they measure different events.

Delivery means the recipient's mail system accepted the message at SMTP level. Your email platform may label this event as delivered, accepted, or successful. It does not necessarily mean the message landed in the primary inbox; the recipient provider may place it in spam, a promotions-style tab, quarantine, or another filtered location.

Deliverability is the broader outcome: your ability to reach the intended recipient's inbox reliably. It is influenced by technical authentication, IP and domain reputation, recipient engagement, complaint behavior, list quality, sending patterns, content, and the receiving provider's own filtering decisions.

This distinction changes how you investigate a problem:

• A spike in hard bounces means mail is not being accepted for some recipients.
• A stable delivery rate combined with lower clicks or conversions may signal inbox placement trouble, weaker content-market fit, tracking issues, or a change in the audience being mailed.
• A rise in spam complaints is an audience and reputation warning even when every message is technically accepted.
• An authentication failure is a configuration and identity problem that can lead to rejection or spam placement.

Gmail explicitly separates delivery errors, spam rate, reputation, and authentication in Postmaster Tools. That is a useful model for your own dashboard: do not collapse distinct failure modes into one vanity percentage. (support.google.com)

The email deliverability metrics dashboard to build

A useful dashboard has three levels: operational health, mailbox-provider signals, and business outcomes. Review them by sending domain, mail stream, campaign type, audience cohort, and recipient provider rather than only as one account-wide average.

1. Operational health metrics

These metrics show whether your sending system can hand mail to recipient systems.

1. Sent volume: messages your platform attempted to send.
2. Accepted or delivered volume: messages accepted by the recipient server.
3. Hard bounces: permanent failures, such as a nonexistent recipient or invalid domain.
4. Soft bounces or deferrals: temporary failures, such as a full mailbox, a temporary provider issue, or a rate limit.
5. Delivery rate: accepted messages divided by attempted messages.
6. Authentication pass rate: the share of messages passing SPF, DKIM, and DMARC, ideally broken out by sending source and recipient provider.
7. Unsubscribe processing time: how quickly your system stops future marketing mail after a request.

2. Recipient-provider and reputation metrics

These describe how mailbox providers and recipients experience your mail.

1. Spam complaint rate: complaints divided by delivered messages, or by the provider-defined denominator when using a provider feedback tool.
2. Inbox placement rate: messages observed in the inbox divided by messages observed in inbox plus spam or junk folders. This normally requires seed-list testing or an inbox-placement provider; a standard delivery webhook cannot establish it.
3. Spam-folder rate: messages observed in spam or junk divided by observed messages.
4. Domain and IP reputation: provider-specific reputation indicators, where available.
5. Provider-specific error rate: rejections and deferrals at Gmail, Yahoo, Microsoft consumer mail, and other meaningful destinations.
6. Engaged-recipient rate: the share of recipients with a recent meaningful interaction, defined consistently for your program.

3. Business and audience-quality metrics

These metrics establish whether inbox access turns into value. They are not direct deliverability measures, but they reveal consequences and protect you from optimizing the wrong thing.

1. Unique click rate and click-to-delivered rate.
2. Conversion rate and revenue or activation per delivered email.
3. Reply rate for conversational B2B, support, sales, or onboarding mail.
4. Unsubscribe rate, separated from spam complaints.
5. Reactivation rate after a win-back sequence.
6. List growth source quality, such as opt-in form, checkout, product signup, event registration, import, or referral.

For product teams sending password resets, receipts, and alerts, delivery errors, latency, authentication, and provider acceptance are usually the leading indicators. For marketers, complaints, inbox placement, engagement by cohort, unsubscribes, and conversion tell the fuller story.

Core formulas and exactly what they mean

Use the same formulas every reporting period. A metric that is imperfect but consistently calculated is more useful than a theoretically ideal metric whose denominator changes from campaign to campaign.

Delivery rate

Delivery rate = accepted messages / attempted messages × 100

Example: if you attempted 100,000 messages and recipient servers accepted 98,700, delivery rate is 98.7%.

Do not calculate delivery rate as sent minus unsubscribes, and do not silently remove bounces from the denominator after the fact. The attempted-message denominator makes changes visible. If your platform automatically suppresses known invalid addresses before sending, report that suppression count separately; it is list hygiene, not a bounce.

Hard-bounce rate

Hard-bounce rate = permanent bounces / attempted messages × 100

A hard bounce should normally result in immediate suppression for that recipient and mail stream, unless the provider gives a credible reason to believe it was a temporary or policy-specific condition. SMTP status codes contain both broad classes and enhanced codes that help software classify delivery failures, but provider text can add important context. RFC 3463 defines enhanced mail-system status codes for richer, machine-readable delivery-status reporting. (datatracker.ietf.org)

Soft-bounce or deferral rate

Deferral rate = temporarily deferred messages / attempted messages × 100

A temporary 4xx-style response is not automatically a lost message. Your provider may retry it. Track both the initial deferral rate and the final outcome after your retry window. A growing deferral rate at one mailbox provider can indicate volume spikes, reputation pressure, infrastructure trouble, or provider-side limits, even if eventual delivery remains high.

Spam complaint rate

Complaint rate = spam complaints / delivered messages × 100

Use the denominator documented by your source when one is supplied. Google states that bulk senders should keep spam rates reported in Postmaster Tools below 0.3%; treat that as a ceiling, not an operating target. Google also requires bulk senders—defined there as those sending more than 5,000 messages per day to personal Gmail accounts—to authenticate mail and provide easy unsubscribe mechanisms. (support.google.com)

A complaint is more serious than an unsubscribe because it tells the mailbox provider the recipient considers the message unwanted. Do not mask the signal by asking recipients to unsubscribe in your email copy instead of making it easy to do so.

Inbox placement rate

Inbox placement rate = inbox placements / (inbox placements + spam placements) × 100

This formula intentionally excludes messages that were not accepted. It asks a different question from delivery rate: among messages that reached the mailbox environment and were measured, where did they appear?

Inbox-placement services use seed addresses across providers. Treat the result as a directional sample, not a census of every subscriber's mailbox. Segment seed results by domain, IP or pool, stream, and content variant. A campaign can perform well at one provider and poorly at another.

Click metrics that remain useful

Unique click rate = unique recipients who clicked / delivered messages × 100

Click-to-open rate = unique recipients who clicked / unique opens × 100

Click-to-open rate can help evaluate content among people whose opens were recorded, but it should not be used alone to judge deliverability. Apple Mail Privacy Protection prevents senders from seeing whether a recipient opened a message and hides IP information, making open data less reliable as a behavioral signal for affected users. (support.apple.com)

Prefer clicks, conversions, replies, account activity, and explicit preference actions when building engagement segments. An email without links can use downstream actions, replies, or a first-party event tied to the campaign as its success signal.

Track by segment or your averages will mislead you

An account-wide rate can look healthy while a high-value audience is receiving poor service. The solution is segmentation that follows how mailbox providers assess mail and how your business sends it.

At minimum, break your metrics down by:

• Recipient provider: Gmail, Yahoo, Outlook/Hotmail, corporate domains, and other large cohorts in your list.
• Sending domain and subdomain: for example, updates.example.com versus notify.example.com.
• Mail stream: transactional, product, newsletter, promotional, lifecycle, and sales outreach should not share one undifferentiated report.
• IP or provider pool: especially if you use dedicated IPs, several providers, or separate sending regions.
• Acquisition source: product opt-in, checkout, content form, co-marketing, event registration, imported list, and so on.
• Recipient age and engagement cohort: new subscribers, recently active subscribers, dormant subscribers, and reactivated subscribers.
• Campaign and template: useful for catching one broken template, deceptive subject line, or unexpected audience selection.

This segmentation supports root-cause analysis. If only marketing mail to Gmail begins deferring, investigate Gmail-specific volume patterns, authentication, headers, complaints, and engagement before changing your receipt template. If hard bounces rise only from an imported source, stop mailing that source and validate it rather than changing your sending infrastructure.

Authentication metrics: measure identity before you chase content

Mailbox providers need a reliable way to associate your messages with your domain. SPF, DKIM, and DMARC are the core controls to configure and monitor. They are not a guarantee of inbox placement, but they establish authenticated identity and are baseline requirements for many senders.

Google requires all senders to Gmail to use SPF or DKIM and requires bulk senders to use SPF, DKIM, and DMARC. For direct mail, Google also requires alignment between the domain in the visible From: header and either the SPF or DKIM domain in order to pass DMARC. (support.google.com)

What to record

For every sending service and domain, monitor:

• SPF pass or fail, including the envelope-from domain.
• DKIM pass or fail, including the signing domain (d=) and selector.
• DMARC pass or fail for the visible From domain.
• Alignment status: whether the authenticated domain aligns with the visible From domain.
• The sending source: provider, IP, application, region, and mail stream.
• Unknown sources appearing in DMARC aggregate reports.

DMARC is an authentication, policy, and reporting protocol. Aggregate reports help domain owners see the sources sending mail that claims to be from their domain and the associated authentication results. (dmarc.org)

Practical DNS examples

The records below are examples, not copy-and-paste production records. Your email service provider supplies its own SPF inclusion domain and DKIM public-key value. A domain must publish only one SPF TXT record; combine authorized mechanisms into that one record rather than adding multiple v=spf1 records.

; SPF example at example.com
example.com. TXT "v=spf1 include:spf.your-email-provider.example -all"

; DKIM example; the public key is supplied by your sending provider
s1._domainkey.example.com. TXT "v=DKIM1; k=rsa; p=PUBLIC_KEY_FROM_PROVIDER"

; DMARC monitoring example
_dmarc.example.com. TXT "v=DMARC1; p=none; rua=mailto:dmarc-reports@example.com; adkim=r; aspf=r"

The p=none example asks receivers to monitor rather than enforce a quarantine or rejection policy. Use it while you inventory every legitimate sender and examine reports. Once all legitimate streams authenticate and align, a stricter policy can protect the domain from unauthorized use. DMARC documentation provides an example record using p=reject, pct=100, and an aggregate-report rua address. (dmarc.org)

A frequent failure is authenticating the parent domain while sending a visible From address from an unaligned subdomain, or enabling a new SaaS tool without giving it DKIM signing and an aligned From domain. Another is moving to p=reject before confirming that support systems, billing tools, CRM sequences, and third-party platforms all pass alignment.

Complaint, unsubscribe, and engagement metrics work together

The healthiest programs make it easier to leave than to complain. Unsubscribes are not necessarily a deliverability failure; they can be a successful preference action that prevents a future spam complaint.

Measure both unsubscribe rate and complaint rate

Unsubscribe rate = unsubscribes / delivered messages × 100

Complaint-to-unsubscribe ratio = spam complaints / unsubscribes

There is no universal “good” ratio because it varies by audience, frequency, acquisition method, and message type. Use it as a trend and segment diagnostic. If unsubscribes remain steady while complaints climb, investigate whether the unsubscribe link is hidden, the one-click mechanism is broken, the audience was not expecting the mail, or targeting became broader.

For marketing and subscribed messages, Yahoo recommends a functioning List-Unsubscribe header, with the RFC 8058 POST method highly recommended, a visible body unsubscribe link, and honoring unsubscribe requests within two days. (senders.yahooinc.com)

One-click unsubscribe headers

A typical implementation looks like this:

List-Unsubscribe: <https://example.com/unsubscribe?token=UNIQUE_TOKEN>
List-Unsubscribe-Post: List-Unsubscribe=One-Click

The first header identifies an unsubscribe endpoint. The second signals that the endpoint supports a one-click POST workflow defined by RFC 8058. Your endpoint must process the request safely without requiring login, and it should suppress the recipient from the relevant marketing stream promptly. RFC 8058 exists partly to avoid accidental unsubscriptions caused by software fetching URLs in headers. (datatracker.ietf.org)

Keep the visible unsubscribe link in the message body even when headers are present. Header support improves mailbox-provider user interfaces; the body link gives recipients a clear, universal route to manage preferences.

Use engagement carefully

Do not define “engaged” as “opened an email” alone. Because some clients obscure open activity, use a layered policy, such as a recent click, purchase, logged-in product activity, reply, or explicitly confirmed preference. The exact window should match your normal buying or usage cycle; a weekly newsletter and an annual renewal program should not use the same recency rule.

For low-engagement cohorts, reduce frequency first, then run a clearly identified re-permission or win-back campaign. Suppress recipients who remain inactive under your documented policy. Do not repeatedly resend promotional mail to people who did not open it; that can worsen complaints and reputation.

Diagnose bounce and deferral codes instead of guessing

A bounce message is evidence, not a generic instruction to “clean the list.” Preserve the SMTP reply, enhanced status code, destination domain, send timestamp, and message stream in your logs.

A practical classification model

Permanent recipient failures usually warrant suppression. Examples include an address that does not exist or an invalid destination domain.

Temporary failures need controlled retries. Examples include a temporary mailbox condition, a provider-side capacity issue, or a transient connection problem. If a temporary error persists over your provider's retry period, review it as a final failure but retain the original reason.

Policy and authentication failures need configuration or sending-practice remediation. Yahoo's sender documentation notes that permanent 553 or 554 errors can result from invalid addresses, DMARC or DKIM policy failures, or message characteristics that are not accepted for policy reasons. (senders.yahooinc.com)

Rate or reputation-related deferrals need volume control and audience narrowing. Avoid reacting by retrying more aggressively. A better response is often to slow the sending rate, send first to recently engaged recipients, and investigate the stream that changed.

A bounce-reporting table

Create a weekly table with these columns:

Before sending an imported or old list, validate addresses and remove obvious risks. An email address verification tool can reduce avoidable invalid-address sends, but verification cannot prove that a person wants your email or guarantee inbox placement. Consent, expectation, and engagement remain separate deliverability inputs.

A worked example: find the cause of a deliverability decline

Assume a SaaS company sends a monthly product newsletter from news.example.com. It mails 80,000 recipients, and the campaign report shows:

Attempted:                 80,000
Accepted:                  79,040
Permanent bounces:            480
Final temporary failures:     480
Spam complaints reported:      72
Unique clicks:              1,185

Calculate the basic metrics:

Delivery rate = 79,040 / 80,000 × 100 = 98.8%
Hard-bounce rate = 480 / 80,000 × 100 = 0.6%
Complaint rate = 72 / 79,040 × 100 ≈ 0.09%
Unique click rate = 1,185 / 79,040 × 100 ≈ 1.50%

At first glance, the 98.8% delivery rate looks acceptable. But the team compares the campaign with the prior three sends and sees that click rate fell, complaints doubled, and nearly all hard bounces came from addresses acquired through one webinar partner.

Next, the team segments the data:

• Gmail complaint and click performance are stable.
• Yahoo traffic has more temporary deferrals than usual.
• Corporate domains account for most hard bounces from the webinar segment.
• DMARC reports show all legitimate marketing sources passing and aligned.
• A seed-list test shows no broad spam-folder problem.

The diagnosis is not “our domain is blocked.” It is a list-quality and expectation problem concentrated in a particular acquisition source, plus a provider-specific deferral issue that needs monitoring.

The recovery plan is concrete:

1. Immediately suppress the hard-bounced webinar addresses.
2. Pause promotional sends to the unengaged portion of that partner-acquired segment.
3. Send a targeted confirmation or preference message only where the original consent language supports it.
4. Keep product and transactional streams separate from the affected marketing cohort.
5. Reduce the Yahoo send rate for the next campaign and watch final delivery after retries.
6. Compare complaints, deferrals, and click-to-delivered rate by provider on every send for the next several campaigns.

How do you know it worked? The answer is not one improved open rate. You should see hard-bounce rate fall for the cleaned segment, deferrals return toward the previous provider-specific pattern, complaint rate stabilize or decline, and clicks or conversions recover among the audience still being mailed. If only delivery rate improves because fewer addresses are attempted, check the cohort size and downstream conversion before claiming success.

Build an operating cadence, not a one-time audit

Deliverability is cumulative. A domain can be correctly authenticated today and still suffer if it repeatedly sends unwanted mail, changes volume abruptly, or mixes reliable transactional messages with low-quality promotions.

Before each major campaign

• Confirm the From domain and DKIM signing domain are aligned.
• Confirm SPF includes the active sending provider.
• Check that the unsubscribe footer, List-Unsubscribe headers, and preference-center links work.
• Review audience rules, exclusions, consent source, and recent engagement.
• Avoid sudden, unexplained volume expansion; phase large sends from the most engaged cohort outward.
• Test rendering and links, including the unsubscribe endpoint.

Weekly

• Review bounces, deferrals, complaints, unsubscribes, clicks, and conversions by recipient provider and stream.
• Review Gmail Postmaster Tools if you send enough Gmail mail to produce data; its dashboards cover reputation, spam rate, authentication, and delivery errors. (support.google.com)
• Process provider feedback-loop complaints into suppression systems. Yahoo's Complaint Feedback Loop is domain-based and requires DKIM-signed email so Yahoo can identify the sender. (senders.yahooinc.com)
• Investigate material changes with a before-and-after view of volume, template, subject line, audience, acquisition source, and infrastructure.

Monthly

• Audit every tool permitted to send as your domain.
• Review DMARC aggregate reports for unknown IPs and failed legitimate senders.
• Review dormant-recipient policies and suppress or re-permission cohorts that no longer demonstrate interest.
• Compare email performance by acquisition source, not just campaign.
• Verify that transactional mail remains on a protected, separately monitored stream.

Your sending provider should expose event webhooks or logs for accepted, bounced, deferred, complained-about, and unsubscribed messages. Check your provider's email API setup documentation for event names and the exact fields available; terminology varies, so map those provider events into your own normalized dashboard rather than assuming identical definitions.

Common mistakes that make metrics look better than reality

Mistake 1: Reporting only delivery rate

A message accepted by the recipient server may be filtered away from the primary inbox. Pair delivery rate with complaint rate, provider errors, engagement trends, and inbox-placement testing where the stakes justify it.

Mistake 2: Treating open rate as proof of interest

Open tracking depends on remote-image behavior and client privacy features. Use opens as a limited directional signal, but make clicks, conversions, replies, product usage, and explicit preferences the stronger decision inputs.

Mistake 3: Hiding poor list quality with a changing denominator

Removing bounces from a delivery-rate calculation after sending makes results look healthier. Report attempted volume, pre-send suppressions, hard bounces, deferrals, and accepted volume separately.

Mistake 4: Combining marketing and transactional mail

A promotional campaign can generate complaints and engagement problems that should not endanger password resets or receipts. Separate domains or subdomains, reporting, templates, and audience policies where your provider and architecture permit it.

Mistake 5: Trying to solve a reputation problem with subject-line tweaks

Copy matters, but persistent deferrals, complaints, or provider policy rejections often trace back to authentication, list source, consent, frequency, targeting, volume changes, or reputation. Diagnose the message stream before rewriting it.

Mistake 6: Treating an unsubscribe as failure

For unwanted promotional mail, an unsubscribe is usually preferable to a complaint. Make the option visible, process it accurately, and offer meaningful preferences if you have more than one relevant mail stream.

What success looks like

A good email deliverability program does not chase a single universal benchmark. It produces stable, explainable trends and makes failures attributable.

You know the program is working when:

• Your sending domains consistently pass SPF, DKIM, and DMARC alignment for every legitimate source.
• Hard bounces are promptly suppressed and are traceable to their source.
• Temporary failures are retried responsibly and monitored by provider.
• Complaint data reaches your suppression logic quickly.
• Unsubscribe paths work in both the email body and supported headers.
• Marketing sends are concentrated on recipients with clear consent and demonstrated interest.
• Provider-specific shifts are visible before they become account-wide problems.
• Transactional and promotional mail can be analyzed and protected independently.
• Email outcomes are measured with clicks, conversions, replies, or product actions—not inflated confidence from opens alone.

The durable goal is not to “beat” a spam filter. It is to maintain an authenticated, recognizable sending identity and send messages recipients expect often enough to be useful, but not so broadly or frequently that they reject them. Your metrics are the early-warning system that makes that discipline operational.

FAQ

What is the most important email deliverability metric?

There is no single sufficient metric. Start with delivery rate, hard-bounce rate, complaint rate, authentication pass rate, provider-specific deferrals, and meaningful engagement or conversion. If you can measure it, add inbox placement rate to distinguish accepted mail from inboxed mail.

Is a 99% delivery rate good?

It indicates that recipient servers accepted most attempted messages, but it does not prove inbox placement or audience satisfaction. Check spam complaints, provider errors, inbox-placement tests, clicks, conversions, and list-segment quality before concluding deliverability is healthy.

What is the difference between a hard bounce and a soft bounce?

A hard bounce is generally a permanent delivery failure, such as a nonexistent address, and should usually be suppressed. A soft bounce or deferral is temporary and may succeed after controlled retries; track both the initial deferral and its final outcome.

Why did my open rate change even though deliverability did not?

Open tracking can change because of client privacy protections, image loading behavior, measurement setup, subject lines, audience mix, or actual inbox placement. Apple Mail Privacy Protection can prevent senders from seeing whether a message was opened, so opens should not be the sole engagement or deliverability metric. (support.apple.com)

Do SPF, DKIM, and DMARC guarantee inbox placement?

No. They verify and align sending identity, which is foundational for deliverability and required by major mailbox providers in many circumstances, but recipient complaints, engagement, list quality, sending behavior, and provider-specific filtering still affect where mail lands. (support.google.com)