Can smooth over the bad moments
Usually the average change between neighboring ping samples. Useful in some packet analysis, but it can hide short high-delay periods that users actually feel.
measurement guide
Why We Use Latency Spread, not Jitter
Bufferbloat.org does not show a generic jitter number because the question we care about is more specific: how far do the high-delay ping samples move away from the median during quiet, download, and upload load? We call that latency spread, and it is one of the signals the bufferbloat test measures.
Run the bufferbloat testSee latency spread on your own connection.
What latency spread is measuring
Latency spread is the gap between the median ping and the 95th percentile ping in a phase of the test. You do not need the notation to read the scorecard, but for completeness the calculation is:
latency spread=P95(ping)-P50(ping)
Median ping is the center of the measured samples. P95 is near the high end: high enough to reflect bad delay moments, but less fragile than using the single worst ping. Latency spread is the distance between those two points.
Another way to read P95: sort the scored pings from lowest to highest, then look near the 95% mark. About 95% of the measured pings are at or below that value, so it represents the upper-delay behavior without handing the result to one isolated spike.
This is deliberately different from showing a vague jitter number. Jitter can mean several different calculations depending on the tool. Latency spread says exactly what is being compared and why it matters for a browser bufferbloat test. If you want the broader context, read how latency under load works.
visual model
Three summaries, three different answers
The same ping samples can be summarized as average jitter, worst ping, or latency spread. Average jitter focuses on sample-to-sample movement. Worst ping focuses on one sample. Latency spread focuses on the upper-delay behavior relative to the median.
Why we do not call this jitter
Jitter is a real networking term, but speed tests often use it as a loose label for several different ideas. Sometimes it means the average change from one ping sample to the next. Sometimes users read the worst ping in a run as if it were the instability number. Those are related signals, but they answer different questions.
Average sample-to-sample jitter can hide the problem we care about: a line can look calm most of the time and still have a few high-delay moments that disrupt calls or games. Worst ping has the opposite problem: it can overreact to one isolated browser pause, Wi-Fi hiccup, or scheduling delay. That is why this page pairs with the separate explanation of why the test uses median ping.
Bufferbloat.org uses the more literal term latency spread because it says what the scorecard is actually showing: how much distance there is between the median ping and the upper end of the measured pings during quiet, download, and upload phases.
Question: which number best captures the high-delay moments users feel without letting one odd ping sample dominate the result?
Can overreact to one sample
Easy to understand, but brittle in a browser. One Wi-Fi retry, browser pause, or scheduling hiccup can become the whole story.
Best fit for this test
Uses p95 minus median. It catches repeated high-delay behavior while avoiding the fragility of grading the run by a single outlier.
How the test calculates it
For each phase, the test records ping samples. It finds the median, which is the center of the sampled pings for that phase. It also finds the 95th percentile, which sits near the high end of the samples without being the single worst outlier. Latency spread is the difference between those two values.
This makes the number practical rather than theatrical. It still sees the uncomfortable upper end of the run, but it does not let one strange browser hiccup define the whole connection. The exact exported field is documented in the technical details, and the measurement flow is described in the methodology.
p95 versus maximum
Why p95 is not the same as worst ping
The maximum ping is the single highest sample. The 95th percentile is still near the high end, but it is less exposed to one-off noise. That is why latency spread uses p95 rather than the maximum: it measures the upper-delay behavior of the run, not just the most dramatic dot on the chart.
How to read it
Smaller is steadier. A low latency spread means the worse moments stayed close to the median ping. A high latency spread means the connection had enough delay swings to feel uneven, especially for calls, low-latency games, and interactive work.
Do not read latency spread alone. The bufferbloat grade still asks the main question: did download or upload load add delay? Latency spread answers the follow-up: even if the median looks acceptable, did the samples stay tight enough for the connection to feel stable? That is also why the result includes an internet connection quality scorecard rather than one isolated number.
The easiest way to see it is to run the bufferbloat test. The scorecard shows the spread next to the measured phases so you can compare what happened while the line was quiet, downloading, and uploading.
check your own line
See whether latency spread shows up on your connection.
Run the test and compare quiet-line ping with what happens while download and upload load are active. The result shows latency spread beside the measured chart, not as a vague jitter number.