D-01 · REPUTATION · IP WARMUP
Shared vs dedicated IP warmup: when to isolate reputation
A new dedicated IP with no history performs worse than a warm shared pool. This scenario documents the volume thresholds, the 30-day ramp, and the idle-IP trap that ruins premature migrations.
Symptom
A SaaS team sending 12,000 emails per month upgrades to a dedicated IP for “better deliverability.” Within a week, Gmail deferrals spike, Microsoft throttles connections, and inbox placement falls below the old shared pool. The team assumes the provider is broken, but the provider is behaving exactly as designed: mailbox providers distrust unfamiliar IPs with sudden volume, and the new address has zero positive history to offset the pattern.
The deeper symptom is variance. Shared pools absorb spikes because aggregate volume stays smooth; a lone dedicated IP exposes every burst. A product launch that doubles weekly sends looks like snowshoe spam from a cold IP. Without a ramp plan, the team oscillates between throttling and overcorrection, warming nothing.
Cause
Mailbox providers score IPs on volume consistency, complaint ratios, bounce cleanliness, and authentication alignment over weeks. A fresh IP starts neutral-to-suspicious, with strict provisional rate limits. Sending full production volume on day one violates the implicit contract: prove predictable, wanted mail first, then scale. Most failed warmups also mix streams — transactional and bulk sharing one IP — so a promotional spike contaminates the reputation that password resets depend on.
The idle-IP trap compounds this. Teams buy dedicated capacity “for later,” send sporadically, and let reputation decay between bursts. An IP that sends 50,000 one week and 2,000 the next three weeks never stabilizes. Consistency matters more than peak volume, which is why providers generally suggest sustained daily volume near six figures before dedicating — lower-volume senders do better on managed shared pools.
Fix
Stay shared until daily volume is steady and placement-sensitive. When migration is justified, split streams first: transactional on one IP or pool, marketing on another, so warmup traffic is homogeneous and engaged. Start with the most engaged segments — recent openers and active trial users — at 5–10% of target daily volume, doubling every few days while deferral and complaint rates stay flat.
Authenticate the sending subdomain fully (SPF, DKIM, DMARC at least at quarantine-monitoring) before day one, and keep content identical to the proven shared-pool templates during warmup. Any creative overhaul mid-warmup confounds the signal: providers cannot distinguish a reputation problem from a content problem. Freeze templates, ramp volume, then iterate.
Prevention
Document the warmup as a runbook: daily caps per provider, engagement-first segment order, rollback thresholds (for example, deferral rate above 5% pauses the ramp for 48 hours). Monitor postmaster tools daily rather than opens. Keep the old shared stream as fallback for two weeks so a warmup stall never blocks transactional mail. Revisit the dedicated decision quarterly — volume that justified isolation can fall back below threshold after seasonality, and returning to shared is a valid outcome.