How to fix email deliverability

By Dana Okafor, who runs Sendnexa's monitoring bench. The sequence below compresses six years of marketing data science into the version that survives contact with real accounts.

Step 1: verify SPF, DKIM, and DMARC actually pass

Not "exist." Pass. Send a test to a Gmail address, open the original message, and read the authentication results line: all three should show pass against your sending domain. The common failures are an SPF record that exceeds its DNS lookup limit after years of vendor additions, and a DKIM signature tied to a domain nobody recognizes because the ESP default was never changed. DMARC belongs at an enforcement policy (quarantine or reject) eventually; p=none is a monitoring stage, not a destination.

Step 2: read Postmaster Tools before changing anything

Google Postmaster Tools shows domain reputation, spam-complaint rate, and authentication trends for mail to Gmail, free. Two numbers matter first: whether complaint rate has crossed 0.1%, and whether domain reputation reads low or bad. If reputation is already damaged, plan for a multi-week recovery; a referenced repair took six weeks to bring complaints from 0.42% to 0.08%, with send volume down 38% during the suppression phase.

Step 3: run a seed test to find out where mail actually lands

ESP dashboards report delivery, which includes the spam folder. A GlockApps seed test sends your real campaign to a controlled panel across Gmail, Outlook, and Yahoo and reports placement per provider. The gap can be dramatic: 99% delivered against 71% inboxed in one baseline we cite. Test before the repair to set the baseline, and again after each major change.

Step 4: find the segments generating complaints

Complaints concentrate. Cross-tab complaint events by segment and send type and the pattern is usually obvious: recipients with no purchase or click in 180+ days, mailed at full campaign frequency. In one audit that shape covered 71% of total send volume. This is where the repair stops being a DNS exercise and becomes a list decision.

Step 5: suppress before you optimize

Pull the complaint-generating segments out of routine sends entirely. Expect the topline dashboard to look worse for two weeks; sends drop immediately while placement gains lag. The coffee-subscription engagement in ourcase-study library cut 22% of volume with revenue flat, because those sends had been producing complaints rather than orders. Suppression is also reversible, which makes it the safest aggressive move available.

Step 6: fix the unsubscribe path

RFC 8058 one-click unsubscribe is required for bulk senders under the 2024 rules, and it is also complaint prevention: a recipient who cannot leave in one click reports spam instead, and a spam report costs reputation where an unsubscribe costs only a row. Honor requests immediately; slow processing shows up as complaints from people who believed they had already left.

Step 7: rebuild volume on engagement, and keep the monitors on

Recover send volume starting with your most engaged segments, watching Postmaster Tools weekly and re-running seed tests as volume steps up. Deliverability is not a one-time fix; it is an operating constraint, which is why ourdeliverability service is monitoring, not a project. Programs that hold complaints under 0.1% get to spend their attention on revenue instead of repair.

What does healthy look like?