Synthetic Monitoring
Continuously executing a real user journey against production from outside it, so a broken journey is detected before a user reports it.
Component health checks answer "is the service up". Synthetic monitoring answers "can someone actually complete the journey", which is a different question and the one that matters. It runs the real path — log in, search, add to basket, check out with a test instrument — on a schedule, from several locations, against production.
It catches a class nothing else does: expired certificates, DNS and CDN misconfiguration, a third party that is failing only for requests from a particular region, an authentication flow broken by a configuration change, and the entire category of failures where every component reports healthy and the journey does not work.
It also gives an availability measurement independent of traffic, which is what lets an SLO be meaningful for a low-volume or overnight-quiet service.
Two requirements. The journey must be genuinely representative, including authentication — a synthetic that skips login misses the most common breakage. And the test's own data must be identifiable and excluded from business metrics and from fulfilment, which is a design task rather than an afterthought.