Data validation in integrations is like a seatbelt. You don't notice it when things are fine. You notice it a lot when things aren't.
The most common integration failures we see aren't technical — they're data quality problems. A field that should be a number arrives as text. A required field is empty. A date is in the wrong format. A postcode doesn't match the state.
Small problems, but they cascade.
What goes wrong without validation
- Invoices with missing line items end up in your accounting system
- Customer records with invalid email addresses break your marketing automation
- Duplicate records from slightly different formatting (John Smith vs JOHN SMITH)
- Financial reporting that doesn't reconcile because amounts arrived as strings
- Downstream systems that crash or produce wrong results from unexpected inputs
The damage isn't just the bad data itself. It's the time spent finding it, fixing it, and figuring out how long it's been wrong. That's where the real cost hides.
Where to validate
The golden rule: validate at the boundary. When data enters your system — whether from an API, a file import, or a user form — check it before processing.
- Input validation: Is the data in the expected format? Are required fields present? Are values within acceptable ranges?
- Referential validation: Does this customer ID actually exist? Does this product code match a real product?
- Business rule validation: Is this order total reasonable? Is this date in the future when it shouldn't be?
Practical validation rules
- Reject records with missing required fields — don't try to guess
- Normalise formats on entry (trim whitespace, standardise case, parse dates)
- Log rejected records with clear reasons so they can be fixed and retried
- Set up alerts when rejection rates spike — that usually means something changed upstream
- Don't validate silently — if you fix data automatically, log what you changed and why
Validation isn't glamorous. But it's the difference between an integration that runs reliably and one that slowly poisons your data without anyone noticing.