Intermediate – Builder · Module A
Diagnosing Customer Integration Failures
A structured approach to "it worked yesterday, it's broken today."
What this session covers
- Reproduce first, theorize second — resist the urge to guess a root cause before you've seen the actual failing request.
- Work outward in layers: is it auth, is it the payload shape, is it network/DNS, is it rate limiting, is it a downstream dependency?
- Compare a known-good request against the failing one field by field — the difference is almost always the answer.
- Always ask 'what changed?' — a deploy, a config change, an expired credential, or a new customer data pattern, in that order of likelihood.
Worked example
A customer's orders stopped syncing at 02:00. Working outward: auth OK (200 on /me), payload — the failing request has `amount: "1,200.00"` while the last good one had `1200.00`. 'What changed?': their ERP upgrade at 01:30 changed number formatting. Root cause found by diffing two requests field by field.
Exercises
- Given a good and a failing JSON request that differ in three subtle ways, find all three differences and say which one would cause a 400.
- Write your own layered checklist (auth → payload → network/DNS → rate limits → dependencies) with one command or check for each layer.