← All work

Case study

Webhook Pipeline for Multi-Tenant Commerce Sync

Multi-tenant platform ingesting commerce events from 10+ external platforms. Polling created 4+ minute lag.

Commerce IntegrationsPlatform Performance / Reliability

Architecture diagram

Webhook-driven commerce event pipeline

Per-platform adapters normalize commerce webhooks into tenant-safe, replayable, near-real-time processing.

Commerce platforms

10+ upstream event sources

Webhook events

>

Webhook adapters

Signature validation and normalization

Normalized payload

>

SQS + DLQ

Backoff, replay, failure isolation

At-least-once delivery

>

Idempotency layer

UUID keys and duplicate protection

Safe event

>

Tenant processors

Ordering and throttling per tenant

Ordered updates

>

Fresh commerce state

Sub-10s data freshness

  • Polling waste was removed instead of simply increasing polling frequency.
  • DLQ and replay made platform outages recoverable instead of silent data loss.
  • Tenant-level throttling kept noisy tenants from affecting shared reliability.

The challenge

Multi-tenant platform ingesting commerce events from 10+ external platforms. Polling-based sync created 4+ minute lag while the system required sub-10s freshness. Lambda costs rising due to inefficient polling.

Architecture approach

  • ·Webhook ingestion layer with per-platform adapters and signature validation
  • ·DLQ + replay queue with exponential backoff
  • ·UUID-based idempotency at ingestion and processing
  • ·Per-tenant ordering via partition keys
  • ·Dynamic throttling with tenant-level rate limits
  • ·Per-tenant metrics, latency percentiles, failure rate dashboards

Tech stack

Node.jsSQSDynamoDBAPI Gateway

Results

  • Data freshness: 4+ minutes → sub-10 seconds
  • Lambda costs dropped ~15%
  • Zero data loss during platform outages