3

We would like to implement slightly different behavior in our web application for those who visit from a certain state (Florida) vs. those in the rest of country (and world).

It's not critical that it be 100% accurate but well into the 90s-- it's OK if a few Floridians see the other version, and others see the Florida version.

What is your experience with state-level identification of users based on their IP address?

Yes I understand the theoretic limitations of using this information (ISPs, proxy servers, yada yada), but I don't know how this plays out in the real world. There are some great answers about the different providers and accuracy, but they don't really answer our state-level empirical question. Google provides some, but not the information we're looking for.

Thanks. We don't want to pursue this approach if it won't work good enough.

Community
  • 1
  • 1
ndp
  • 21,546
  • 5
  • 36
  • 52
  • Daniel's answer below is excellent. I'm curious if anyone has first-hand experience trying to rely on this data (although we all rely upon for ad serving) for other things. – ndp Aug 24 '10 at 03:50
  • I did rely on MaxMind's data, but only on a country level. We used it for a credit card fraud prevention system. The system used to display to the human operators the client's country derived from the IP address (from Maxmind), the country of the bank where the card used was issued (from the payment gateway) and also the country that the user chose when creating the account. If any one of these didn't match, the operator would take manual action (phone the client, or similar)... – Daniel Vassallo Aug 24 '10 at 07:34
  • ... The country derived from the IP address was regarded as the most reliable of the three, and even though I did no scientific sampling, I believe that false/missing/incorrect responses were much less than 98%... I know you require state level accuracy, but from the good experience on a country level, I guess the stats published by MaxMind won't be much further from the truth... Of course, you may want to wait and see if someone has better first-hand experience on this level. – Daniel Vassallo Aug 24 '10 at 07:36

1 Answers1

1

MaxMind, which offers a popular IP-to-location database published some statistics on its data:

This is the data they published for USA (last updated May 17th, 2010):

  • Correctly Resolved Within 25 Miles of True Location: 83% (85%)

  • Incorrectly Resolved More Than 25 Miles From True Location: 14%

  • Not Covered on a City Level: 3%

Definitions:

  • Correctly Resolved Within 25 Miles of True Location: refers to the IP addresses that are correctly resolved within 25 miles of the actual city. The second number in parenthesis refers to the percentage of IP addresses that are correctly resolved if excluding data not covered on a city level.

  • Incorrectly Resolved More Than 25 Miles From True Location: refers to IP addresses that were incorrectly resolved by more than 25 miles of the actual city.

  • Not Covered on a City Level: refers to the percentage of the data of select country that does not contain city information. We leave these data ranges without city information to prevent false positives. The city information is not available for AOL IP address or when Internet Service Providers (ISPs) route traffic through proxies thus masking the originating location.

Community
  • 1
  • 1
Daniel Vassallo
  • 337,827
  • 72
  • 505
  • 443