
What Does an API Integration Cost? (And Why Estimates Vary 10x)
A simple one-way sync with a clean, modern API runs €2,000–€6,000. Bidirectional sync with field mapping and conflict rules: €8,000–€20,000. A legacy system with no real API — file drops, screen scraping, SOAP archaeology — €20,000–€50,000+. The spread isn't padding: it's error handling, auth lifecycles, and two departments discovering they define "customer" differently.
"Just connect A to B" is the most innocent-sounding request in software. Here's why the quotes for it can legitimately differ by a factor of ten.
The three tiers of integration cost
| Tier | What it looks like | Typical cost |
|---|---|---|
| One-way sync, modern API | REST/webhooks, one direction, few fields | €2,000–€6,000 |
| Bidirectional sync | Both systems write; mapping, dedupe, conflict rules | €8,000–€20,000 |
| Legacy without a real API | File drops, screen scraping, SOAP archaeology | €20,000–€50,000+ |
Why do estimates vary 10x?
Because the happy path is maybe 20% of the work. Two systems exchanging data when everything works is a week. The rest of the budget goes to what happens when things don't:
- Errors and retries. The other side will be down, slow, or return something undocumented. You need retries with backoff, idempotency, and dead-letter queues.
- Partial failures. 950 of 1,000 records synced — what happens to the other 50? Silent data loss is how integrations destroy trust.
- Rate limits. Fine at 100 records, throttled at 100,000.
- Auth lifecycles. Tokens expire, credentials rotate, and someone revokes access on a Friday afternoon.
- The data-mapping workshop. The consistently underestimated part: sales and accounting discover that "customer" means two different things, and someone has to decide which system wins each field. That's an organizational decision wearing a technical costume.
Questions that make your quote accurate
Answer these before asking for a price. Which direction does data flow — one way or both? What happens on conflict — which system wins? How fresh must the data be — real time or nightly? What volume — hundreds or millions of records? Does the target system have a documented, supported API, and can we get test credentials today? Sharp answers here compress a 10x range into a real number.
APIs change — budget for it
An integration isn't done at go-live. The other side deprecates versions, renames fields, and tightens rate limits. Plan a small annual budget for keeping the connection alive — far cheaper than the emergency rebuild after a surprise deprecation notice.
Building integrations is core to our process automation services — and if you're weighing tools before writing code, read Make vs. n8n vs. custom.

