Latency vs the World Average: How the Comparison Works
GlobeLatency does not judge your latency with a fixed millisecond limit. It divides your average for each destination by that destination's world average, collected from many visitors. Up to 1.15 times the world average is good, up to 1.5 is slightly slower, and above that is clearly slower, with safeguards so tiny absolute differences never look like problems.
On this page
Quick answer: every destination in GlobeLatency has a world average: the typical latency visitors get to that exact server. Your result is divided by it. A ratio of 1.0 means you are exactly typical. Below 1.0 you are faster, above 1.0 you are slower. This keeps the verdict fair, because a far server and a near server are each judged against what is normal for that server.
Why a relative comparison?
Latency is dominated by distance. Light in fibre travels roughly 200,000 km per second, so a round trip to a server 10,000 km away cannot be quicker than about 100 ms even on a perfect route, and real routes are longer than straight lines (see undersea cables and the speed of light). If a tool used one fixed rule such as "over 100 ms is bad", then everything far away would look broken, and your result would depend mostly on geography rather than on the quality of your connection.
A relative comparison asks a better question: "Is my connection to this server better or worse than what other people normally get?" That is the question that matters when you wonder whether your ISP, your Wi-Fi or your route has a problem.
How the world numbers are built
Each time someone completes a Test, their per-destination averages are recorded. The server keeps, for each destination and each day, the average, the lowest and the highest latency observed. Combined across days these become the world average, world low and world high that you see in each tile. Because they come from real users on mobile, Wi-Fi, fibre and cable, the world average represents an ordinary mixed audience rather than an ideal lab setup.
Custom destinations that you add yourself have no world data, since they are never stored. Their tiles show your numbers only and use the fixed fallback bands.
The ratio and its bands
| Your average ÷ world average | Meaning | Colour |
|---|---|---|
| 0.70 or less | Great: at least 30% faster than typical | Green |
| 0.70 to 1.15 | Good: typical, or up to 15% slower | Green |
| 1.15 to 1.50 | A little slower than usual | Yellow |
| Above 1.50 | Clearly slower than usual | Red |
The percentage on the tile is the same number written differently: ratio 1.20 is +20%, ratio 0.60 is -40%.
The two safeguards
A pure ratio would be unfair at small values. If the world average is 12 ms and you measure 20 ms, the ratio is 1.67, which sounds terrible, yet 8 ms is imperceptible. So a tile is never red when:
- the difference from the world average is less than 30 ms, or
- your value is under 50 ms.
These safeguards stop trivial differences from creating alarming colours, and they match how latency is felt by people: below roughly 50 ms nearly everything feels instant.
The overall verdict
The headline verdict uses the same logic on the whole test. Your overall average is compared with the world average of the same destinations you tested. Results map to Excellent, Good, A little slower than usual or Laggy. If nothing replied, the verdict is No replies.
The comparison bar
Under the summary you see a bar with the world average at the centre. Your position to the left means faster than the world; to the right means slower. It is a quick visual of the same percentage shown on the card.
Worked examples
Example 1: far server
World average 210 ms. You measure 185 ms. Ratio 185 / 210 = 0.88, which is Good and about 12% faster. In absolute terms 185 ms is high, but for this far server it is better than typical.
Example 2: near server
World average 24 ms. You measure 60 ms. Ratio 2.5, the value is above 50 ms and the difference (36 ms) is above 30 ms, so no safeguard applies and the tile is red. Something is wrong on your side for this route: perhaps a VPN or a poor ISP path.
Example 3: safeguard in action
World average 14 ms. You measure 33 ms. Ratio 2.36, but the value is under 50 ms and the difference is 19 ms, so the tile stays green or yellow instead of red.
Example 4: no world data
You add a private hostname. No world average exists, so the tile uses the fixed bands: 38 ms is great, 90 ms is good, 200 ms yellow, 300 ms red.
Limitations of the world average
- It is based on GlobeLatency visitors, who may be tilted towards certain countries or tech-savvy users.
- It mixes access types. If you are on satellite or mobile, you will often be slower than a fibre-heavy average, and that is a true observation about your link.
- Destinations that changed hosting or routing recently take a few days to settle in the statistics.
- A destination with very few tests has a noisy average.
For the wider picture see Accuracy and limitations.
Common mistakes and myths
- "Green means my internet is fast." Green means typical or better for those servers. Throughput (speed) is a different thing; see latency vs bandwidth.
- "Red means my ISP is bad." One red tile may be a route issue to that one destination. Several red tiles across regions is more telling.
- "I should always beat the average." About half of users are slower than average by definition.
What to do next
If most tiles are yellow or red, follow troubleshoot high latency. If you want absolute guidance by use case, see what is a good latency. Or run a test now and compare.
Frequently asked questions
Where does the world average come from?
It comes from latency tests run by visitors of GlobeLatency. Measured averages, lows and highs for each destination are stored per day and combined into a global statistic for that destination. It reflects the audience of the site, not a perfectly balanced sample of the entire planet.
Why not use fixed limits like under 50 ms is good?
Because the same number means different things for different servers. 150 ms to a server on the other side of the world is excellent, while 150 ms to a server in your own city is poor. Comparing against the same destination removes distance from the judgement.
What if a destination has no world data?
GlobeLatency falls back to fixed bands: under 50 ms great, under 120 ms good, under 250 ms slightly slow, above that slow. This is the only case where absolute numbers decide the colour.
Is being 10 percent slower than the world average bad?
No. Up to 15 percent above the world average is shown as good, because normal variation between networks, times of day and devices is larger than that.