Multi-channel lifecycle engine

Customer follow-up was manual and inconsistent. An earlier scheduled version fired unpredictably and had no way to tell a delivery failure from a send that never happened — so nobody could say whether a campaign had actually reached anyone.

• 01

  Rebuilt the scheduler around a fixed two-hour window with explicit day-offset routing, replacing the cadence logic that had been misfiring

• 02

  Joined every send against a suppression list and an already-sent ledger before dispatch, so a retry can never double-message a customer

• 03

  Throttled dispatch to one message at a time with a fixed delay, keeping the send rate under the provider's frequency ceiling

• 04

  Logged failures as a first-class path — every rejection stored with its provider error code, not collapsed into a generic error

• 99.6% email delivery across the measured window (1,123 of 1,127)

• Throughput grew roughly 18× in four weeks while the failure rate fell from 28% to 2%

• Every failure is now attributable to a specific cause — provider frequency capping accounts for the majority, which is a tuning problem rather than a defect

• 21 campaigns run across four lifecycle journeys with no manual intervention

• n8n
• MySQL
• WhatsApp Cloud API
• SendGrid
• GCP
• Docker