Observability
The property of being able to answer new questions about a system's internal state from its external outputs, without shipping new code.
The distinction from monitoring is worth holding onto. Monitoring answers questions you predicted: is CPU high, is the error rate above threshold. Observability answers the question you did not predict: why are requests from this one tenant, on this one API version, slow only after a deployment?
What makes that possible is high-cardinality, high-dimensional data — events carrying tenant, version, region, endpoint, user agent and more — rather than pre-aggregated counters, because aggregation discards exactly the dimension you turn out to need.
Logs, metrics and traces are the usual three pillars, but the useful framing is not the pillar count. It is whether an unanticipated question can be answered in minutes from data already being collected.