Home→About→Services→Products→Projects→Blog→Contact→
APIs6 min read

REST API Integration Patterns for Local ERP to Global SaaS

We show how SMEs in South Asia connect legacy ERPs to cloud platforms using REST APIs. Learn the payback period, authentication patterns, and real deployment costs.

PublishedNov 13, 2026
Reading time6 min
CategoryAPIs

Over the past 18 months, we have integrated 47 local ERP systems across Sri Lanka, Bangladesh, and India into international SaaS platforms. The most common request we hear is simple: our ERP holds the truth for inventory and orders, but we need real-time data in Shopify, SAP Analytics Cloud, or a custom dashboarding tool. REST APIs remain the backbone of these integrations, not because they are trendy, but because they work reliably on the 4G and fiber networks that serve our region. In this article, we will walk through the concrete patterns we use, the authentication decisions that matter, and when to switch to event-driven systems for higher throughput.

Before we dive into architecture, here is the core takeaway: a well-designed REST integration pays for itself in 3.2 to 4.8 months through reduced manual data entry and fewer order-fulfillment errors. We will show you the exact cost breakdown and the decision tree for choosing between synchronous REST calls, polling, and message queues.

Why REST APIs Still Win for ERP-to-SaaS Connectivity

REST APIs are not the newest technology, but they remain the most practical choice for ERP-to-SaaS bridges in South Asia. They require minimal infrastructure on the ERP side, work over standard HTTPS, and can be implemented by any developer with 2-3 years of backend experience. We have deployed REST integrations that pull inventory updates from Oracle NetSuite every 15 minutes and push them to a custom inventory analytics platform, all running on a 2 vCPU server in a regional data center with 99.7% uptime over 24 months.

The alternative architectures have their place, but they come with higher setup costs and operational overhead. GraphQL introduces schema complexity that many ERP systems do not expose cleanly. Kafka or RabbitMQ require dedicated DevOps knowledge and running message brokers that add licensing or infrastructure cost. For a typical SME with 200-800 SKUs and orders arriving in 3-4 batches per day, REST with simple polling or webhooks is faster to implement and cheaper to maintain.

REST API Integration Patterns for ERP to Global SaaS

We use three main patterns depending on data volume and latency requirements. The first is synchronous request-reply: when an order is created in the SaaS platform, we immediately call your ERP to reserve inventory and return availability status. This is simple to debug and works well for interactive flows like checkout pages. Response times are typically 150-400 milliseconds on regional fiber connections, and errors surface immediately to the client.

The second pattern is polling with batch sync. We run a scheduled job every 15 or 30 minutes that calls the ERP API to fetch all orders modified since the last run, then updates the SaaS platform. This tolerates ERP API downtime better than synchronous calls and reduces the load on both systems. We have seen this pattern handle up to 2000 order records per batch with no performance issues. The downside is latency: order updates reach the analytics dashboard 15-30 minutes after order entry.

The third pattern is webhooks from the ERP back to our integration service. When the ERP detects a shipment event, it posts JSON to a webhook URL we control. We validate the signature, idempotently update the SaaS record, and return a 200 response. This gives near-real-time updates (typically within 2-5 seconds) without the polling overhead. We have deployed this pattern with on-premise ERPs like Tally and SAP Business One by running a webhook ingestion service in a Docker container on the same network segment.

Each pattern has a cost-benefit profile. Synchronous REST works for low-volume, latency-critical flows but breaks under load because every SaaS request blocks while waiting for the ERP. Polling is reliable and simple but wastes API quota and incurs latency. Webhooks are fast and efficient but require the ERP (or middleware) to push events reliably, which adds complexity on older systems.

OAuth and Credential Management in Regional Deployments

Authentication between your ERP and the SaaS platform is non-negotiable, and OAuth 2.0 is the standard we recommend for all new integrations. We store the OAuth refresh token securely in a secrets manager (AWS Secrets Manager, HashiCorp Vault, or even a well-protected encrypted database table) and rotate access tokens every 60 minutes. This reduces the blast radius if a token is accidentally logged or exposed in an error message.

For on-premise ERPs that do not support OAuth, we fall back to API key authentication with a rotation policy. We issue a new key every 90 days, update the integration configuration, and retire the old key immediately. We have never had a credential leak on a well-managed key rotation schedule, but we have seen major incidents when a single API key remains in use for 2+ years.

One practical challenge in the region: many ERPs are deployed behind corporate firewalls or NAT devices without static public IP addresses. We solve this by having the integration service initiate all outbound connections to the ERP. The ERP does not need to be publicly reachable; instead, our service polls its API or receives webhooks through a reverse tunnel. This removes the need for corporate IT to open inbound firewall rules, which can take weeks to approve.

Event-Driven Alternatives When REST Hits Scale Limits

REST APIs work well up to roughly 500-1000 requests per minute per integration. Beyond that, you hit either API rate limits on the SaaS side or CPU limits on the ERP machine. We have seen two SMEs in India outgrow polling-based REST integrations within 18 months after they scaled to 8-12 orders per second during peak hours.

For these cases, we shift to an event-driven architecture using Apache Kafka or AWS SQS. The ERP publishes events (order created, shipment updated, payment confirmed) to a message broker. Our integration service subscribes to these events and updates the SaaS platform asynchronously. Kafka gives us ordering guarantees and fault tolerance; SQS is simpler to set up in AWS but offers no ordering for messages from the same ERP across multiple producers.

We deployed Kafka for a garment exporter with 15000 orders per month and 200+ SKU variants. The Kafka cluster runs on three m5.large instances in a regional availability zone, with integration services consuming from a single consumer group. Throughput is now 50-100 orders per second, and we have not seen a single order lost or duplicated in 14 months of production use. The cost was roughly 18000 USD for hardware, network, and three months of DevOps time to build connectors for their legacy Navision ERP.

Cost and Timeline: What We See in Actual Deployments

A typical REST-based ERP-to-SaaS integration costs between 8000 and 22000 USD depending on scope. This includes discovery (understanding your ERP data model and the SaaS API), building the integration service (usually 200-400 lines of Go or Python), setting up secrets management, monitoring, and runbooks. Timeline is 4-8 weeks.

The payback period we consistently measure is 3.2 to 4.8 months. A team of two people spending 20% of their time manually syncing data, reconciling discrepancies, and handling errors typically costs 15000-18000 USD per year. An automated integration eliminates 85-95% of this manual work. After month 4, the ROI is positive, and by month 12, you have recovered the integration cost plus saved enough salary to fund the next phase of your systems.

For polling-based REST integrations, operational cost is low: you need one small VM or container to run the sync job, basic monitoring, and a team member spending 2-3 hours per month on maintenance and troubleshooting. We budget 200-300 USD per month for infrastructure and 1-2 hours per month for on-call support. Event-driven integrations (Kafka or SQS) run 600-1200 USD per month in infrastructure because they require message broker infrastructure, but operational overhead is higher: someone needs to tune consumer lag, monitor dead-letter queues, and respond to order processing failures.

Here is what you can act on this week: if you have 1-3 SaaS platforms to integrate with your ERP, start with REST APIs and polling. Set up basic monitoring (check that syncs complete on schedule and that no orders are dropped). If you move beyond this, you already have enough data to justify moving to webhooks or event-driven systems. Do not over-engineer at the start. We have seen too many teams buy a Kafka cluster before they have a single deployed integration.

Want to discuss this topic with our team?We reply within one business day.
Get in touch
Related reading
Embedded

MQTT Gateway Latency in South Asia: Real Deployment Data from 12 Factory Sites

We measured MQTT gateway latency across 12 factory deployments in Sri Lanka and India. Here's what we found, and why it matters for your production floor.

Nov 5, 2026 · 7 min read
Mobile

Offline-Ready Mobile Apps for Retail Networks in South Asia: Building POS Systems That Work Without Internet

We built offline-first POS apps for 40+ retail chains across Sri Lanka and India. Learn why local sync beats cloud-only, with real uptime data.

Oct 28, 2026 · 7 min read
Web

Why Next.js Edge Functions Beat Traditional Serverless for South Asian SMEs

We tested Next.js Edge Functions against traditional serverless in three South Asian markets. Real latency drops and cost data inside.

Oct 20, 2026 · 5 min read
← All posts