measurement guide
What a bufferbloat speed test measures
A speed test usually tells you how much data can move. A bufferbloat test asks the next question: does the connection stay usable while that data is moving? Bufferbloat.org measures capacity, quiet-line delay, delay while the line is busy, and how far the high-delay samples move away from normal.
The short version
A throughput-only speed test can say the line is large. A quiet ping test can say the line is quick when almost nothing is happening. A bufferbloat test checks the behavior people actually feel: whether the connection keeps responding while download and upload traffic are active.
The signals in the scorecard
- Download throughputhow much data the connection can receive during the download load phase.
- Upload throughputhow much data the connection can send during the upload load phase.
- Quiet-line latency / pingthe normal response time before the test adds download or upload traffic.
- Latency during download loadresponse time while the browser is receiving data and the downstream path is busy.
- Latency during upload loadresponse time while the browser is sending data and the upstream path is busy.
- P95 spreadthe 95th percentile ping minus the median ping during each phase. This is the upper-delay spread shown on the scorecard instead of a generic jitter number.
What turns those numbers into a bufferbloat result
The core bufferbloat signal is the difference between quiet-line latency and loaded latency. If ping stays close to normal while the test is downloading and uploading, the connection is behaving well under pressure. If ping rises sharply, queues are probably adding delay.
Throughput still matters for some uses, especially streaming and large transfers, but low throughput alone is not bufferbloat. A slower connection can still feel good if latency stays stable. A fast connection can feel bad if latency jumps whenever the line is busy.
Why p95 spread is included
Two connections can have the same median ping and still feel different. The one with a lower upper-end spread is usually easier for calls, games, and remote work to handle. That is why the scorecard includes p95 spread as a supporting signal for application performance.
Bufferbloat.org uses the plain term latency spread because browser tests and network tools may calculate jitter differently. The current method uses 95th percentile ping minus median ping so the scorecard captures bad-but-representative moments without treating one isolated spike as the whole result. The reasoning is explained in why we use latency spread, not jitter.
What this test does not measure
This is a short browser test, not a full network audit. It does not prove long-term uptime, diagnose Wi-Fi interference by itself, map every ISP routing problem, or replace sustained packet-loss testing. If a result looks surprising, run the test again and compare.
Why the test is open
Bufferbloat is easy to miss because it can hide behind good-looking speed-test numbers. A test built to expose it should be inspectable: the code, methodology, and limitations should be public so the result can be checked, debated, and improved.
See what happens when your own line gets busy.
Run the browser test to compare quiet-line ping with download and upload load. The scorecard shows the measured trace, application performance, and exportable technical details.