Throughput: download and upload

Download throughput is the rate at which your device receives useful data. It affects page assets, software downloads, streaming and incoming video. Upload throughput is the rate at which your device sends data, including camera video, backups, source files and screen sharing.

Both are normally reported in megabits per second. They are not fixed properties of a place: radio conditions, protocol overhead, server capacity, traffic management and simultaneous users all influence the observation. A short burst may also differ from a sustained transfer.

Latency: how long an interaction takes to travel

Latency measures delay. A round-trip measurement includes the journey from a device to a test endpoint and back. Physical distance, routing, radio scheduling, queuing and network processing all contribute.

Low latency makes conversations feel natural and remote controls respond promptly. High latency does not necessarily prevent a large download, because many packets can be in flight at once. This explains why a network can achieve a strong Mbps result yet feel slow in an interactive tool.

Jitter and packet loss: consistency matters

Jitter is a practical name for variation in packet delay. Real-time applications use buffers to absorb some variation, but large or irregular changes can produce pauses and robotic audio. Reporting only the average latency can hide this behavior.

Packet loss means expected data did not arrive. Reliable transfers can retransmit it, reducing effective throughput. Real-time streams may conceal or skip missing information instead. Even a small percentage can be noticeable when loss occurs in bursts.

Technical references

Stability and availability are derived evidence

Stability summarizes whether repeated observations remain within a useful range. It must state how many observations, over what duration and which metrics were considered. A stability score derived from one short session should not be described as long-term reliability.

Availability can describe coverage, successful test completion or the probability that a service is usable. These are different definitions. GCI publishes the exact role of availability in its methodology and avoids inventing a current availability value when the source pipeline cannot support it.

How to read a result in context

QuestionWhat it revealsWhat it does not prove
When and where?Relevance to your intended usePerformance at every venue or time
Wired, Wi-Fi or mobile?The access condition testedThe quality of other access methods
How many samples?Breadth of the evidenceRepresentativeness by itself
Which endpoint?Part of the network pathPerformance to every service
Idle or busy?Whether contention was exercisedLong-term capacity without repetition
Which data period?Freshness and comparabilityThat the methodology stayed unchanged

Why GCI uses an evidence-labelled composite

A traveler needs a simple comparison, but simplicity creates risk. GCI’s Mobile Traveler Score combines component metrics while keeping data mode, period, sample size and confidence available. Baseline, hybrid and measured are not interchangeable labels.

Use the score to sort and explore. Use the components and evidence labels to decide. For integration work, the proposed Connectivity API & Data Pro keeps those fields attached so a product cannot accidentally present a dated estimate as a live measurement.