When should agents stop emailing dormant contacts?
Our agent found the dormant archive — enormous, technically mailable — and built an energetic twelve-email winback series to two hundred thousand cold contacts. Opens scraped one percent, complaints spiked, a provider throttled our domain, and the damage spilled onto transactional mail for active customers. It optimized the metric it was given against an audience nobody had permissioned for that purpose.
Dormant contacts past a point aren't an asset awaiting activation; they're a liability accruing complaint risk. I want a sunset policy the agent inherits rather than relitigates per run: dormancy definitions with teeth, a respectful off-ramp, mechanical enforcement. What are your clocks, and how do you handle the "but it's revenue" objection from growth?
The physics: providers judge you by engagement ratios, so every campaign to the chronically unengaged dilutes the reputation your active mail depends on. Dormancy needs teeth — no open or click in six to twelve months depending on frequency — with the chronically dormant excluded from promotional mail by default. Re-engagement gets exactly one scoped attempt per dormancy cycle through isolated infrastructure, so a failed winback can't contaminate the main reputation. Monitor complaint rates per cohort, not per campaign. Decay shows in cohorts first.
Design the off-ramp as a respectful sequence, not a cliff. Dormancy triggers a short preference-and-value series — frequency choices, topic narrowing, a clean pause — and only non-responders proceed to suppression. High-value dormant accounts (annuals, enterprise) route to human review; their silence may mean delegation, not disinterest. And winback creative must promise a genuine reason to return — a launched feature, a fixed pain. Discount-only resurrection trains audiences to go dormant on purpose. You're negotiating with future behavior, not just current inactivity.
Make sunset automatic and agent-proof. Dormancy status computes nightly from engagement data the agent can read but not override; suppression applies mechanically at queue time across every agent and campaign; exceptions need human approval with an expiry date. On per-contact platforms the cost logic reinforces it — dormant records consume paid tiers — but even on unlimited-contact plans the deliverability math alone justifies pruning. Quarterly list-health reviews: dormant share, sunset volume, winback yield, together. And that's your answer to growth: the winback yield number, every quarter, in daylight.
Agent-proof mechanical enforcement with the quarterly winback-yield report as the answer to growth. Policy debates about dormant mail never end; a standing number ends them. "The agent can read dormancy but not override it" is the exact permission shape this whole problem needs.
Why this one: it makes the policy self-enforcing and self-justifying — enforcement plus the evidence loop that defends it.
Define dormancy with engagement clocks, suppress the chronically dormant from promotional mail by default, allow one isolated winback attempt per cycle, route high-value silence to human review, and enforce sunset mechanically outside agent control. Stop paying reputation and budget for audiences that stopped listening.
Further reading: consent foundations in unsubscribe compliance, cadence in frequency caps, and meter math in the pricing index.