All IP to location solutions are inaccurate. 51Degrees LocationConfidence gives you an edge.
No IP intelligence solution, including 51Degrees, is going to locate the exact latitude and longitude of a device in practice. Carrier Grade Network Address Translation (CGNAT) and other technologies are not only used by cellular ISPs but also residential ones.
This is particularly true in Europe, and increasingly so in the United States.
Digital Element, IPInfo, MaxMind, and 51Degrees are all providing the best guess based on the source data used to train the algorithm.
The differences relate to the training data and the algorithm. 51Degrees obsesses over both.
We know what we don’t know and keep you in the know. 😉
Classification
51Degrees solution classifies the location returned into three confidence bands which must be checked before using the returned values.
- High confidence results can be relied upon and considered useful for all practical purposes.
- Medium confidence results might be useful where the use case can tolerate a degree of inaccuracy.
- Low or Unknown confidence results should be discarded when the results might be used to make consequential decisions. Instead of using the results, consider asking the user to share their location via device APIs such as GPS and then using reverse geocoding, or enter their postal address manually. Subsequent processing may also be similar to handling an exception when obtaining the result.
Comparison
Let’s take a look at what this means in practice using data from a website using CloudFlare on 19th November 2025. We believe CloudFlare are using IPInfo based on publicity.
The Difference column in the following table shows the average difference in kilometres between the latitude and longitude returned from 51Degrees and from CloudFlare. The IP Count column shows the number of unique IP addresses included in the sample. The columns are grouped by the 51Degrees confidence in the result.
| Confidence | ||||||
|---|---|---|---|---|---|---|
| High | Low | Medium | ||||
| Continents | Difference | IP Count | Difference | IP Count | Difference | IP Count |
| Africa | 67 | 2,038 | 1,281 | 126 | 287 | 133 |
| Asia | 47 | 13,934 | 1,960 | 9,576 | 191 | 1,890 |
| Europe | 88 | 83,589 | 487 | 15,577 | 510 | 4,558 |
| North America | 92 | 47,595 | 3,156 | 11,816 | 486 | 7,413 |
| Oceania | 61 | 2,733 | 1,583 | 248 | 694 | 214 |
| South America | 72 | 10,991 | 627 | 528 | 266 | 576 |
| Average / Total | 84 | 160,880 | 1,704 | 37,871 | 448 | 14,784 |
The average difference for 51Degrees high confidence results is 84 kms, for medium 448 kms, and for low 1,704 kms. The differences vary by continent.
75% of the IP addresses in the overall sample have a high 51Degrees confidence. The result can be relied upon.
7% are classified as Medium confidence.
Most significantly 18% of the results are low confidence, meaning the location information should be ignored.
Depending on the use case for the IP location, and its sensitivity to accurate location information, 18% of the traffic being received might be incorrectly treated.
For analytics this will result in noise where the incorrect geography is being considered.
For advertising this might mean wasted advertising spend.
What are you waiting for?
Take Location Confidence from 51Degrees for a trial alongside your existing solution.
- Always include the high confidence ones.
- Play around with medium confidence results.
- Ignore the low confidence results.
Then see what happens to your business outcomes.
- Do you make more money?
- Do you eliminate noise, making better decisions?
- Do bounce rates reduce as users aren’t misled with incorrect location information?
Contact us now to get started.