Point-to-Point vs. Middleware-Based Integration
The first architectural decision is whether to connect Salesforce directly to your ERP via REST or SOAP APIs, or to introduce a middleware layer such as MuleSoft, Dell Boomi, or Workato. Point-to-point integrations are faster to build for a single connection but create exponential complexity as endpoint count grows. Middleware provides message transformation, routing, error handling, and monitoring as shared infrastructure.
- Point-to-point suits single-ERP environments with fewer than five integration touchpoints
- Middleware becomes essential when connecting Salesforce to multiple ERP instances or legacy systems
- Salesforce Platform Events and Change Data Capture enable event-driven architectures without polling
- API call governor limits on Salesforce (100,000 per 24 hours on Enterprise Edition) require batching strategies
- Middleware provides canonical data models that decouple Salesforce object schemas from ERP table structures
Synchronous vs. Asynchronous Data Flow Patterns
Real-time integrations using synchronous callouts from Salesforce Apex triggers provide immediate consistency but introduce latency and failure coupling. Asynchronous patterns using Platform Events, Outbound Messages, or middleware queues offer better resilience at the cost of eventual consistency. Most production architectures combine both: synchronous for critical validations and asynchronous for bulk data movement.
- Salesforce Outbound Messages trigger SOAP callbacks to middleware on record creation or update
- Platform Events publish to an event bus that external subscribers consume with CometD or Pub/Sub API
- Change Data Capture streams field-level changes on Account, Opportunity, and custom objects in near-real-time
- Queueable Apex and Batch Apex handle high-volume asynchronous callouts within governor limits
- Idempotent message handling in the ERP prevents duplicate order creation during retry scenarios
Error Handling and Monitoring Strategy
Integration failures between Salesforce and ERP systems are inevitable. Network timeouts, data validation rejections, and schema drift all produce errors that must be detected, logged, and resolved before they impact business processes. A robust error handling strategy includes dead-letter queues, automated retry policies, and dashboards that surface integration health in real time.
- Dead-letter queues in MuleSoft or Boomi capture failed messages for manual review and resubmission
- Salesforce Custom Objects store integration error logs with links to the originating Account or Order record
- Exponential backoff retry policies prevent thundering herd effects during ERP system outages
- Monitoring dashboards track message throughput, latency percentiles, and error rates per integration flow
- Alerting via Slack or PagerDuty triggers when error rates exceed thresholds or queues back up beyond SLA
Try it here, free
Manufacturing Process Types and Product-Process Matrix Reference
Compare project, job shop, batch, cell, repetitive, assembly line and continuous processes by volume, variety, layout, planning approach and typical ERP setup.
Ready to design a Salesforce-ERP integration architecture that scales? Contact our integration architects for a discovery session.