02 · Software · infrastructure · 2026
Cold email engine
The machine that turns a client's market into sending campaigns: pull the audience, verify every address, split it into cells, write the copy, size the inboxes, and load it all paused for a person to launch.
The decision
Cells, not campaigns
Early on, each client had dozens of campaigns named however the day went, and nobody could say which one was winning or why. I replaced them with a grid. Every campaign is a cell named client, timezone, email provider and number, capped at 600 leads. The timezone decides when it sends, the provider decides which inboxes send it, and the cap means a bad list can only hurt one cell. For one automotive client that is 36 cells and 18,371 leads, each readable at a glance.[48] Sourceengineering notes (workspace grid rebuild), one automotive clientAs of2026-09-21StatusVerifiedAll sources →
What I built
- The pipeline end to end: audience from the warehouse, verification bands, suppression of the client's own customers, the cell grid, copy, and the load into the sequencer.
- Copy loading that never goes live by itself: an approved plan is written into the sequencer paused, read back, and compared by hash against what was approved.
- Infrastructure sizing as a margin equation: 20 sends per inbox per day, ramp counted as half-volume, and inbox spend held under 10% of the contract.[50] Sourceengineering notes (infra spend model), live fleet constantAs of2026-08-07StatusVerifiedAll sources →
- Deliverability rules in code: weekdays only, subject lines varied with true synonyms but never on the client's brand, catch-all addresses held back until the verifier can vouch for them.
What was hard
A syntax that fails silently
Our sequencer only honours one form of spintax. The common form is sent to prospects literally, braces and all, and nothing errors. I found it in a live campaign, then made the linter refuse the wrong form before anything loads.
Throughput that looked fine for weeks
Dividing by calendar days instead of sending days sized fleets at well under half of what they needed, and the shortfall only showed up as throttled campaigns a month later. The sizing now counts sending days and ramp explicitly.
What I learned
- Name things so the name answers the question you will ask at 1am.
- The most useful number is often the inverse one: not how many inboxes a list needs, but how big a list the budget can serve.
Results
- The system behind the client results below, delivered as approved by each client.