Instacart's engineering team describes how they built a scalable multi-tenant marketing platform to support 350+ white-label retailers on Storefront Pro. The architecture provisions isolated third-party vendor workspaces per retailer, uses a stream consumer with batching (up to 50 users per call) to improve throughput, and routes messages through a Rails/Sidekiq-based CRM Service. Key challenges solved include cross-retailer data isolation, API rate-limit management, IP warming automation, and self-service template customization via a React UI with Liquid templates deployed through CI/CD. The platform achieves 99.9% delivery success and enables retailers to launch campaigns without engineering involvement. Future plans include adaptive campaign optimization and AI-assisted content generation.
Table of contents
TL;DRBackground: Marketing Across Marketplace and StorefrontArchitecture OverviewThird Party APIKey Components1. Instacart Tooling for Retailers2. Campaigns Engine3. Stream Consumer & Batching Layer4. Instacart CRM Service5. Third-Party Provider WorkspacesSolving Scale ChallengesOperational ConsiderationsDomain ConfigurationGet Brent Scheibelhut ’s stories in your inboxIP WarmingObservabilityResilience and RecoveryTemplate Tooling for RetailersContent Creation at ScaleSelf-Service Template CustomizationResults and ImpactAI-Driven Optimization and What’s NextAdaptive Campaign OptimizationAI-Assisted Content GenerationMulti-Channel IntelligenceConclusionAcknowledgmentsSort: