Cold Sending Stack Built on Instantly, Smartlead or Lemlist

The ceiling is the design: Twelve to eighteen mailboxes across four to six secondary domains, each capped at 25 sends a day and never raised past 40. That cap is the design decision, not a limitation of the build. A mailbox at 40 sends a day with real replies coming back is an address providers learn to trust. The same mailbox at 150 is a complaint generator that takes its domain down with it, and takes a month of pipeline with the domain. DNS, done once and written down: SPF kept inside the ten-lookup limit by flattening the include chain and stripping ESP includes the secondary domains never use. DKIM at 2048 bits with a separate selector per sending platform, so changing platform later does not mean rotating keys across everything. DMARC published at p=none with aggregate reports going to a parser I read weekly, then moved to quarantine once two clean weeks are on record. MX on Google Workspace or Microsoft 365, chosen by where the reply will be answered rather than by seat price. Every secondary domain gets a one-page site and a 301 to your main domain: a From: address whose domain resolves to nothing is the easiest decision a filter makes all day. Inside Instantly, Smartlead or Lemlist: A custom tracking domain per sending domain, with open tracking off. An open pixel buys a metric you cannot act on and an HTML wrapper that changes how your message looks to a filter. Link tracking stays off through message two and comes on at message three, where a click carries information. One-click List-Unsubscribe headers (RFC 8058) on every send including the first, which most cold platforms leave off by default and which Google and Yahoo have been explicit about expecting. Sending windows follow the recipient's working hours, and Friday afternoon is off for Dutch segments for the obvious reason. Suppression before anything sends: Deduplication against your CRM: open opportunities, closed-lost inside twelve months, active customers, everyone who unsubscribed from marketing, and every address in your support desk. Built as a sync with HubSpot or Pipedrive where one exists, otherwise as a dated export with a refresh procedure. Under GDPR that file is also your record of objection, so it lives in your storage as a file you own rather than inside a tool you may leave next year. Warm-up, and the day I switch it off: A 21 to 28 day ramp, then reciprocal warm-up pools go off. Those pools trade opens and replies between accounts inside the same network. The behaviour mailbox providers weight is behaviour from recipients they can attribute, and a pool cannot supply that. Leaving warm-up running permanently also hides the only signal worth watching, which is what placement does when real volume goes up. Not included, and not negotiable: I do not source, buy or broker contact data. I do not configure cold sending on the domain carrying your invoices, password resets or portal notifications, and not on a subdomain of it either. I will not build a stack for a list somebody bought: the infrastructure will be correct and it will still burn, and the money is better spent on the list. Handover: A runbook your team operates without me: per-mailbox caps, the ramp schedule, every DNS record with its TTL and where it lives, the response to a 5.7.1 policy block, the response to a 4.7.x throttle, and the three thresholds that mean stop sending today. Two working sessions with whoever will own it. Who this is not for: Teams with nobody named to answer the inbox. A reply to cold outbound has a useful life measured in hours.