The rate at which useful application‑level payload data (excluding protocol headers, retransmissions, duplicates and control overhead) is successfully delivered to the application layer over a communication path, measured over a specified observation interval and usually expressed in bits per second or bytes per second.

Principle

Goodput is always less than or equal to throughput; it quantifies effective user‑level data delivery and therefore directly governs application performance and user experience independently of lower‑layer raw bitrates or control traffic.

Demonstration

Illustrative scenario — Situation: A file transfer application sends payload over TCP on a lossy link. Recognition: Over a measured interval compute the total application payload bytes received, excluding TCP/IP headers and retransmitted payload bytes. Action: Report goodput as payload bytes/sec. Consequence: Retransmissions, large headers, encryption overhead or small packetization can substantially reduce goodput relative to the raw PHY bitrate, lowering perceived transfer speed.

Misapplication

Using link or physical layer bitrates (throughput or nominal bandwidth) as a proxy for application performance without accounting for protocol overhead, retransmissions, encryption and framing, or measuring goodput at a point that still includes duplicates or unstripped headers.

Consequence

Goodput is the appropriate metric for application capacity planning, user experience assessment and billing models that charge for usable data; misconstruing goodput as equal to raw throughput can cause misguided optimization or incorrect SLAs.

Reversal

In some designs, accepting lower goodput (for example by sending stronger headers, redundant control frames, or FEC) intentionally improves reliability, reduces application‑level errors, or lowers latency jitter; thus maximizing goodput is not always the correct objective.

Boundary

Clearly within: Application‑level payload bytes successfully delivered per unit time after stripping protocol overhead and excluding retransmitted payload. Boundary case: Measurements at intermediate protocol layers that strip some but not all headers require clear definition. Clearly outside: Physical layer bitrates and throughput measures that include headers and duplicates are not goodput.

Semantic Tension

Goodput ↔ Reliability/Latency: designs that maximize goodput (e.g., by disabling redundancy) can reduce reliability; conversely, redundancy that improves reliability may reduce goodput—tradeoffs must be resolved according to application needs.

Synthesis

Goodput is the user‑relevant measure of effective payload delivery; it must be measured at the application boundary (or equivalently defined point) and interpreted together with latency and reliability requirements to guide system design.