Guides8 min read

When to escalate a deliverability problem

Not every blip deserves a war room. A simple four-quadrant matrix tells you whether to ignore, monitor, fix, or escalate.

Most deliverability anxiety comes from operators who've seen one bad number and don't know whether to panic or ignore it. A reply rate dips from 4% to 3%, a single seed address landed in spam, Postmaster shows Medium instead of High. Is this an emergency? A ticket? A note in the quarterly review? This matrix gives you a decision rule so you stop asking.

The matrix in one line

Plot the problem on two axes: severity (how much it hurts the business) and trajectory (is it getting worse?). The quadrant tells you the response.

Axis 1: Severity

Severity is the business impact, not the technical interestingness. A DKIM warning on a subdomain you don't send from is zero severity no matter how alarming it looks in a dashboard. A 40% drop in transactional delivery is critical severity even if everything on the reputation dashboard is green.

  • Low: single metric mildly off, no revenue impact, no user complaints
  • Medium: noticeable metric shift, indirect revenue impact, isolated user complaints
  • High: multiple metrics off, direct revenue impact, active user complaints, regulatory exposure
  • Critical: emails not reaching users who need them (password resets, order confirmations, payment receipts), major provider block, legal exposure

Axis 2: Trajectory

Trajectory is about the past two weeks. Is the problem getting worse, staying flat, or recovering? A metric stable at a new low for 14 days is a different beast from a metric dropping 5% per day.

  • Improving: the number is moving back toward healthy
  • Flat: no movement for two weeks
  • Slow decline: losing 1–3% per week
  • Cliff: losing 5%+ per week or single-event drop over 20%

The four quadrants

Quadrant 1: Low severity + improving/flat — ignore

The metric is mildly off but stable or recovering. Don't spend time on it. File a one-line note in the quarterly review doc and move on. Common examples: one seed mailbox landing in spam on one test, Postmaster reputation ticking from High to Medium on a week with a holiday campaign, complaint rate at 0.05% (half the threshold) holding steady.

Quadrant 2: Low/medium severity + slow decline — monitor

The metric is trending the wrong way but hasn't crossed any important threshold yet. Set up a weekly check, log the number, and give yourself four weeks before acting. Usually the trend stabilises on its own, or you catch it early enough to fix with a cheap intervention (remove inactive subscribers, slow volume growth).

Quadrant 3: Medium/high severity + flat or slow decline — internal fix

Real problem, not yet a fire. This is your internal project territory. Assign an owner, give them two weeks, check in twice. Common examples: DMARC reports showing a vendor sending without DKIM alignment, Microsoft junk folder rate climbing from 5% to 15% over a month, a specific campaign type consistently landing in Promotions.

Before starting the internal fix, run a fresh inbox placement test so you have a before-number to compare against the after-number.

Quadrant 4: High/critical severity or cliff — escalate

Something is actively broken and the trajectory confirms it. This is the quadrant where you wake people up. External escalation means one or more of: a specialist consultant retainer, direct outreach to an ISP postmaster, blocklist delisting process, legal review if there's a DMARC spoofing incident. Common examples: Microsoft blocking all mail with 550 5.7.1, password reset delivery dropping below 80%, Gmail spam rate over 0.3% for three consecutive days, a domain landing on Spamhaus SBL.

Before escalating

Have three things ready: the problem statement in one sentence, the metric timeline (what changed when), and the impact quantified in dollars or users. Escalation without these three is just panic with a bigger audience.

Special cases that jump quadrants

A few scenarios override the matrix because they have second-order consequences the numbers don't yet show:

  • Transactional email failing. Password resets, OTPs, receipts. Even 5% failure is critical severity because users lock out of accounts and contact support. Jump straight to Quadrant 4.
  • Domain on a major blocklist. Spamhaus, Barracuda, UCEProtect L1. Even if your numbers look fine today, the blocklist will catch up with you. Quadrant 4.
  • DMARC spoofing reports. Someone is sending mail as you. Not technically a deliverability problem, but bleeds into one fast. Quadrant 4.
  • Regulatory exposure. GDPR, CAN-SPAM, CASL, 152-FZ complaint from a user. Legal, not deliverability — but often surfaces here first. Quadrant 4 with legal in the loop.

Who gets called in each quadrant

Escalation is about who gets pulled away from what they're doing. The matrix should map cleanly to a staffing decision:

  • Ignore: nobody — it goes in the doc
  • Monitor: the email owner spends 10 minutes a week
  • Internal fix: one engineer/marketer, 4–8 hours over 2 weeks
  • Escalate: consultant or ISP direct contact, could be a week of multiple people's time

If your instinct is to put everything in Quadrant 4 because "email is important", you'll burn your team out and miss the actual emergencies. If your instinct is always Quadrant 1 because "it's just email", you'll discover the problem when a customer calls support about a missing invoice. The matrix forces honesty.

FAQ

How do I tell a real cliff from a normal fluctuation?

Cliff = drop of 20%+ in one day, or 5%+ per week for two weeks. Anything less is noise and should be treated as such. Check the cause list (did you just migrate ESPs, rebrand, run a risky campaign?) before reacting.

What counts as 'critical severity' beyond transactional?

Anything with a direct user-experience or legal implication. Two-factor codes. Shipping notifications. Regulatory notices. Mailing-list unsubscribe confirmations (yes — failure to confirm an unsub can be a CAN-SPAM issue).

How long should I stay in 'monitor' before acting?

Four weeks is the right maximum. If the decline hasn't stopped by then, promote to internal fix. Staying in monitor indefinitely is how small problems become expensive ones.

Can the matrix be automated?

The thresholds can be, and should be. Set up alerts at the boundaries (spam rate crossing 0.3%, placement dropping under 70%, bounce rate over 2%) so the system tells you when a quadrant changes instead of you checking manually.
Related reading

Check your deliverability across 20+ providers

Gmail, Outlook, Yahoo, Mail.ru, Yandex, GMX, ProtonMail and more. Real inbox screenshots, SPF/DKIM/DMARC, spam engine verdicts. Free, no signup.

Run Free Test →

Unlimited tests · 20+ seed mailboxes · Live results · No account required