Building an ERP Integration Portfolio Project That Reads as Real Experience

Why simulate an ERP integration

ERP syncs are the canonical FDE project: legacy constraints, dirty master data, money on the line. You can't access a real ERP ethically - so build the other side as a fixture. Done right, this project reads as delivery experience because it exercises the whole arc.

Architecture

[fixture ERP]  ←REST, 12% error rate, duplicate webhooks→  [sync service]
                                                     ↓
                                            [Postgres: orders, ledger]
                                                     ↓
                                     [reconciliation report + DLQ UI]

The fixture (week 1)

• Container with endpoints: orders, products, customers.
• Deliberate mess: duplicate SKUs with different case, customers missing VAT IDs, order IDs reused after cancellation, a webhook that fires twice 20% of the time, an endpoint that times out under load.
• Include a changelog: "v1.2 added a field" - so you can handle schema evolution.

The sync service (weeks 2-3)

• Watermark-based incremental pull (don't trust webhooks alone).
• Idempotency keys = ERP order ID + version; unique constraint at DB.
• Retry with backoff + budget; DLQ with reasons; quarantine for business-rule violations (missing VAT).
• Reconciliation report: counts in/processed/quarantined/failed, diffable by day.

The delivery story (week 4)

• Runbook: how to tell sync is broken, first 5 checks, how to re-drive the DLQ.
• Deployment story: Docker, config via env, backup of the DB, rollback procedure.
• The honest close: what broke during your own testing (something will), what you'd harden next.

Writing it up

Frame as a customer engagement: brief with success metrics ("100% of valid orders within 15 min, zero silent drops"), then the arc. This mirrors the delivery artifacts in our public capstone brief; its rubric is for self-review, not external grading.

Continue: Portfolio ideas · Production checklist