Data Latency

Latency is the elapsed time between something happening in the world and a consumer being able to act on the data about it. It accumulates at every stage: collection, processing, and delivery each add milliseconds, and the number that matters is the end-to-end figure, not any single hop. Practitioners usually care about the tail as much as the median, because it is the slow 1% of updates that breaks a strategy or a downstream job, not the average.

Reducing latency means moving work toward the source: doing normalization and scoring inline instead of in a downstream batch, and measuring against event time rather than the wall-clock moment a record happened to land. In the fastest corners of markets, firms co-locate hardware next to an exchange and count latency in microseconds; for most alternative-data use cases, the meaningful unit is seconds, and the win is simply not waiting for a nightly cycle.