Why ERP integration with Salesforce fails and how to fix it

SAP and Salesforce integrations usually fail for business reasons, not technical ones. The sync runs, but the data is too old for the decision the user is making, the two systems disagree about which customer record is right, or nobody owns the flow when an update breaks it. The fix is to design from the user's decision backwards, clean the data first, pick middleware that fits your scale, and give every flow an owner and a failure plan.

This piece is for CIOs, sales operations and integration leads connecting Salesforce to SAP. It covers the four failure patterns, the integration options, what a reliable integration needs, and a specification template.

I've worked with businesses where even a one-hour lag between systems caused quoting errors that led to lost deals. The usual cause is that nobody asked how fresh the stock data had to be for a rep to quote from it.

An integration can be technically working and commercially broken. That distinction is where to start.

Wrong sync frequency for the decision

Batch sync is fine for some data. It is not fine for available-to-promise stock, pricing or order status where customers expect an answer now.

The failure pattern: the integration team builds what is technically convenient for the volumes and sets batch intervals on technical grounds. Nobody on the business side checks whether that frequency supports the decision the user is making.

Ask one question before anything is specified: what is the oldest the data can be and still be useful? Daily may be fine for customer master updates. For stock availability in B2B sales, anything over fifteen minutes is a problem.

1. 01Customer masterSAP to Salesforce, 24 hours
2. 02Price listsSAP to Salesforce, 1 hour
3. 03Stock availabilitySAP to Salesforce, 15 minutes
4. 04Won opportunitySalesforce to SAP, immediate
5. 05Order statusSAP to Salesforce, 15 minutes
6. 06Invoice and paymentSAP to Salesforce, 24 hours

Quote, order and invoice agree

Data quality mismatches

If the company name in Salesforce is formatted differently from the customer master in SAP, every sync creates mismatches that need manual fixing.

I've worked with teams that spent hours every week reconciling basic customer information between Salesforce and the ERP, and I've seen projects where just mapping customer hierarchies took weeks because the two systems defined an "account" differently. The systems had been maintained separately for years and had drifted apart in ways nobody had mapped.

Clean the data before you integrate. It sounds obvious. It is the most consistently skipped step.

Wrong architecture for the scale

Point-to-point integration works for two systems, a few flows and stable processes. It is the cheapest way to start.

It doesn't scale. Add an SAP module, a new Salesforce business unit or a third system, and the connections multiply. Each one has to be maintained, tested and troubleshot separately.

Middleware (SAP Integration Suite, MuleSoft, Boomi or similar) gives you one governed layer with central monitoring and error handling. The trade-off is upfront investment in architecture and governance. Companies that deploy middleware but give it no named owner end up with the same problems as point-to-point, plus a platform nobody understands.

No integration governance

Every Salesforce release and every SAP support package can break a flow: API changes, new validation rules, changed fields, new authentication requirements. Integrations that are built and forgotten fail on the first upgrade nobody regression-tested. It is the most common post-go-live failure I see.

SAP Integration Suite. SAP's integration platform on SAP BTP, and the natural choice in SAP-centric landscapes. It handles IDocs, BAPIs, OData and SAP message formats well and has prebuilt content for common SAP scenarios. Complex business processes still need custom integration flows and real integration skills. If you are still on SAP PI/PO, note that its mainstream maintenance ends at the end of 2027, so new Salesforce flows should not be built there. On RISE, check what SAP BTP entitlement your contract already includes before buying more.

MuleSoft. Owned by Salesforce since 2018, with a broad connector library, an official SAP S/4HANA connector and accelerator templates for SAP order-to-cash. A good fit if you already own it. The catch for SAP-heavy organisations: MuleSoft is a separate skill set, and the team running your SAP programme is probably not the team that should design your MuleSoft architecture.

Boomi. A cloud integration platform (independent of Dell since 2021) with SAP and Salesforce connectors and a lower entry barrier than MuleSoft. Reasonable for mid-market organisations that want governed middleware without MuleSoft's cost or complexity.

Point-to-point APIs. Direct REST or SOAP calls between Salesforce and SAP avoid middleware cost. They need disciplined API versioning, regression testing on every release and a team that understands both systems. Fine for simple, stable cases. A debt accumulator for anything complex.

My SAP CPI and Integration Suite guide covers the SAP-side platform in more depth, and the five CRM options for SAP compares the CRMs themselves.

1. A specification of every flow: fields, direction, frequency, key mapping, and what happens when the two systems disagree.
2. An explicit failure design. When a sync fails, what happens to in-flight data? How many retries? Who gets the alert? What is the manual recovery? Most integrations are under-designed here.
3. Regression tests before every upgrade. Salesforce ships three major releases a year, and SAP ships support packages and updates of its own. An automated suite covering the critical flows should run before each one.
4. Monitoring with alerts. A silent failure is worse than a loud one. Errors that pile up for days are much harder to fix than errors caught in minutes.
5. An owner. One person who knows what the integration does, sees when it breaks, and has the access and authority to fix it.

The same discipline applies beyond Salesforce. In my Middle East defence case study, connecting Microsoft Dynamics, Experlogix CPQ and SAP SD cut quote turnaround by 35%.

Fill in one row per flow before anyone writes a specification document. If a cell is blank, the flow isn't ready to build. The values below are illustrations; agree your own with the people who use the data.

Flow	Direction	Trigger and frequency	Oldest acceptable data	System of record	On failure	Owner
Customer master	SAP to Salesforce	On change	24 hours	SAP	Retry, then alert data steward	Customer master data owner
Stock availability	SAP to Salesforce	On request or near real time	15 minutes	SAP	Show "check with operations" flag	Supply chain systems lead
Price lists and conditions	SAP to Salesforce	On change	1 hour	SAP	Block quote on stale price	Pricing manager
Won opportunity to sales order	Salesforce to SAP	On close	Immediate	Salesforce until order created, then SAP	Queue and alert sales ops	Sales operations lead
Order and delivery status	SAP to Salesforce	On each milestone	15 minutes	SAP	Retry, then alert integration support	Integration lead
Invoice and payment status	SAP to Salesforce	Daily	24 hours	SAP	Alert finance systems	Finance systems lead

The "oldest acceptable data" column is the one most teams never fill in. Agree it with the users who make the decisions, not with the integration team alone. If you are already running late, SAP Integration Suite delivery delays covers the common causes.

Customer records don't match. Cause: the two customer masters were never aligned. Fix: reconcile before go-live, make SAP the system of record for customer data, and enforce the mapping in the integration.

Order status not updating in Salesforce. Cause: the integration covers quote-to-order but not the status callbacks. Fix: send status updates from SAP SD back to Salesforce at each order milestone.

Quote prices differ from invoice prices. Cause: prices are maintained by hand in Salesforce and drift from SAP pricing conditions. Fix: make SAP the price master and pull prices into Salesforce through the integration.

Integration breaks after an SAP update. Cause: no integration regression testing in the upgrade plan. Fix: put integration regression testing in the scope of every SAP update.

Can Salesforce integrate with SAP?

Yes, through SAP Integration Suite, MuleSoft, Boomi, other integration platforms or direct APIs. Typical flows are customer master sync, won opportunity to sales order, order and delivery status back to Salesforce, pricing into Salesforce for quotes, and invoice and payment status. A customer master sync is straightforward. Full quote-to-cash with complex pricing, many company codes and real-time stock is a significant project.

Is Salesforce an ERP or a CRM?

A CRM. It manages pipeline, opportunities, customer interactions, marketing and service. It doesn't post accounting entries or manage inventory. SAP is the ERP for finance, procurement, inventory and production. Integrated well, a sales rep sees stock and payment status in Salesforce and finance sees deal value in SAP.

What is the main reason ERP-CRM integrations fail?

Requirements defined from the technology instead of from the user's decision. The integration gets built correctly to a specification that was wrong: data four hours old, prices that don't match invoices, status updates that lag. Before writing a technical specification, document the decisions each user group makes with the data and how fresh it has to be.

How do you maintain an ERP-Salesforce integration over time?

Run automated regression tests on the critical flows before every Salesforce release and every SAP update. Monitor every critical flow with alerts that fire on failure, not in an end-of-day report. Name an owner who has access to the monitoring and is part of change planning for both systems.

When should you use SAP Integration Suite for Salesforce integration?

When SAP is the dominant system, when you have SAP BTP access, and when flows involve SAP-specific content such as IDocs, BAPIs or SAP message formats. It is a weaker choice if you have already invested heavily in MuleSoft or Boomi, if your team lacks SAP integration skills, or if SAP is a minor participant in mostly non-SAP flows. For new RISE programmes with Salesforce in scope, it is the natural starting point.

How long does an ERP-Salesforce integration project take?

A standard scope (customer master sync, quote-to-order and basic order status) using prebuilt content typically takes 8 to 16 weeks from scoping to go-live, assuming clean data and a dedicated integration developer. Full quote-to-cash with complex pricing, multiple entities, credit management and real-time stock typically takes 4 to 9 months. Data problems found mid-project and flows added during build are the usual causes of overrun, so run a data quality assessment before build starts.