Batching

Batching is when the tracking library groups several events into one packet and sends them in a single network request instead of firing each event separately.

The point of batching is thrift: every network request costs battery, bandwidth and connection overhead, so folding ten events into one packet is far cheaper than ten separate calls. But that deliberate delay has a side effect an analyst should know: events do not reach the server at the instant they happen, they arrive slightly later and sometimes out of order, which is why insert time differs from event time. The common trap is that if the user closes the page before the packet is sent, those events can be lost, so good libraries flush the packet on page unload using a beacon mechanism. Batching should not be confused with buffering: the buffer is where events wait, and batching is the decision about when to drain it.