Throughput measures how much work a system processes per unit of time. For a blockchain, the unit might be transactions, bytes of data or computational work measured in gas. The unit and measurement stage must be stated for the number to be meaningful.
Rate is different from waiting time
If a network includes 600 transactions over 60 seconds, its observed inclusion throughput is 10 transactions per second for that interval. This does not mean each transaction waited one tenth of a second. Individual delays depend on submission time, selection, processing and the confirmation standard being used.
Workloads change the result
Simple transfers and complex contract interactions demand different resources. A system may process fewer transactions while doing more computation. Batching can also place many user actions inside one transaction, so counting transactions is not the same as counting payments or users.
Capacity versus actual activity
Observed throughput reflects the work actually processed; it is not automatically the network's maximum sustainable capacity. A quiet network can have spare capacity, while a short test may overstate what is sustainable. Compare equivalent workloads and time windows, and distinguish submitted requests, included transactions and finalized outcomes.