MCP

Throttling and rotation

TL;DR

  • Volume scales horizontally. More mailboxes and more sending domains, each sending a little — never one mailbox sending a lot.
  • Three to five sending inboxes is the usual starting point, with sends rotated across them.
  • Pace the sends across the day instead of firing the whole batch at once. This part is convention rather than published evidence — treated as such below.

Why rotation exists

Reputation attaches to the sending domain and the individual mailbox, so the safe way to double volume is to double the number of senders rather than the load on each. A vendor guide puts the typical starting configuration at three to five inboxes, "because rotating across multiple mailboxes lets you scale safely without overloading any single one" (2026) [1]. Sending platforms implement this as an automatic rotation: the campaign draws from a pool of connected mailboxes and distributes the day's sends across them.

Two consequences worth internalizing. First, the per-mailbox ceiling from Volume limits and ramp applies to the sum of everything that mailbox sends, across every campaign — rotation shares the load, it does not raise the limit. Second, each new mailbox arrives with no reputation and needs its own ramp, so capacity is something you provision weeks ahead of needing it, not on the morning you want to launch.

Sending domains follow the same logic. Cold outreach is normally sent from secondary domains that resemble the primary one rather than from the company's main domain, so that a reputation problem cannot take down the email everyone in the business depends on. Note the tension: the blocklist operator's view names fleets of throwaway lookalike domains as a listing criterion. A small number of clearly owned, properly authenticated secondaries is normal practice; churning through disposable ones is the pattern that gets sending infrastructure burned.

Pacing within the day

The standard advice is to spread a mailbox's daily allocation across its sending window with irregular gaps, rather than dispatching the batch in one burst — the reasoning being that human mailboxes do not emit fifty messages in ninety seconds. It is plausible and universally repeated. It is also, as far as this library could verify, unsupported by any published measurement: the guides that recommend randomized delays do not show the data behind them, and the vendor page most often cited for it does not actually discuss send spacing at all.

So: pace your sends, because the cost is zero and the mechanism is believable, but do not treat delay randomization as a lever that will fix a placement problem. The things that demonstrably move placement are list quality, complaint rate and volume discipline, and no send-timing jitter compensates for those.

Practical shape

  • Provision three to five mailboxes across two or more sending domains before you need the volume, and ramp each one.
  • Let the platform rotate; do not hand-assign prospects to mailboxes.
  • Budget the per-mailbox ceiling once, across all campaigns.
  • Spread the day's sends over hours, not minutes.
  • Add capacity by adding senders — the moment you find yourself raising a per-mailbox limit to hit a number, you are trading a durable asset for a week's pipeline.

References

  1. Smartlead — How to set up cold email (vendor guide, Jul 2026)