Most ERP integrations fail. Not "underperform" — genuinely fail. 40–60% run over budget, 30% miss their original goals entirely, and 15% get quietly written off after 18 months of effort.
The failure modes are consistent. Here are the top 5, and the discipline that separates the projects that ship.
Failure 1: Nobody defined the source of truth
ERP integrations are about connecting two systems that have overlapping data. Customer records live in the CRM AND the ERP. Product data lives in the ecommerce platform AND the ERP. Order data lives in three places.
Without a defined source of truth for each entity — "the CRM owns customer records, the ERP is the read-only copy" — you get sync loops, duplicate data, and finger-pointing when numbers don't match.
Fix: in week 1, draw a data map. For each entity (customer, product, order, invoice), one system is the writer, others are readers. Document it. Sign it.
Failure 2: Sync logic without failure handling
Every ERP integration will have sync failures. Network blips, API rate limits, malformed records, invalid data types. If your integration doesn't have explicit failure handling, you'll silently drop data.
Fix: every sync operation must have:
- Retry logic with exponential backoff
- Dead-letter queue for records that fail repeatedly
- Alerts to a real human (not "look at the logs sometime")
- Reconciliation reports run daily to catch drift
Failure 3: Underestimating the ERP's quirks
Big ERPs (SAP, Oracle, NetSuite, MS Dynamics) have decades of design decisions baked in. Their APIs are shaped by internal domain models that don't match modern REST expectations.
Common surprises:
- Rate limits far below what documentation suggests
- Field validation rules undocumented
- Business logic embedded in workflows you didn't know existed
- "Custom fields" behaving differently than "standard fields"
Fix: budget 20-30% of engineering time for "the ERP surprised us." Don't quote a tight fixed price if you haven't done a full API discovery week first.
Failure 4: Not planning for data migration
ERP integrations usually require migrating historical data — customers, orders, product catalog, financial history. This is often 40-60% of total project effort but gets estimated as "we'll figure that out."
Fix: plan migration as its own workstream:
- Extract from source system (usually export + custom scripts)
- Clean up (dedup, standardise formats, fix bad data)
- Validate (compare totals, spot-check records)
- Load to target system (in stages, not all at once)
- Verify (compare pre/post totals)
- Roll back plan (in case verification fails)
Failure 5: No ops handover
The integration goes live. Six months later, something breaks. Nobody remembers how it works. The engineer who built it left the agency. You're paying someone to reverse-engineer your own system.
Fix: ops handover is a deliverable:
- Architecture doc explaining data flow
- Runbook for common failure modes
- Access to all monitoring dashboards
- 90-day post-launch support window
The discipline that works
For ERP integrations to succeed, you need:
- Discovery week upfront. Full API audit. Data mapping. Business rules documented. Written scope. No fixed price until this is done.
- 2-week sprints with real deliverables. Not "we're 60% done." Actual working integration paths, one per sprint.
- Reconciliation reports from day 1. Not something to add later.
- A dedicated ops person on the client side who owns the integration after handover.
- Realistic timeline. ERP integrations are 12–36 weeks. Anyone quoting 6 weeks is either lying or scoping wrong.
What we do differently
We refuse fixed-price ERP integrations without a discovery week. We ship reconciliation reports before we ship sync logic. We require a client-side ops owner as part of the contract.
If you've been burned by an ERP integration or are about to start one, book a consultation. We'll help you scope it properly whether or not you engage us.