Sometimes in life you need to take things with a pinch of salt. If you are a stickler for the government’s nanny state guidelines of 6 grams per day, then you should steer well clear of Geo/IP targeting!
Firstly, let’s look at some disclaimers:
Please note that geo-targeting accuracy is not guaranteed and may vary depending on various factors, including the level of targeting selected.
Quova states that their IP geolocation data is 99.9% accurate at the country level and up to 98.2% accurate at the US state level.
No other solution provides the accuracy, data depth or feature-richness of NetAcuity.
With country-level (IP country) targeting that’s more than 99.9 percent accurate and city-level (IP city) targeting that’s more than 95 percent accurate, you will increase the effectiveness of online advertisements, and maximize reach, relevance, and revenues.
So it appears some solutions give broad disclaimers, others give absolute accuracy figures and Digital Envoy just claim to be more accurate than everyone else. The honest answer from all of them would be something like:
Geo-Targeting allows you to pretty accurately deliver and report to specific regions with high statistical significance. Some regions will be falsely labeled, resulting in either incorrect delivery or incorrect reporting. The range of error is low enough to achieve campaign uplift.
So how does it work?
Publishers, Networks, Adservers and Ad Verification companies all offer some kind of Geo-Targeting/Reporting. Most buy IP look-up tables from providers like Digital Envoy and Quova. These providers collate lists of IP addresses and their associated Geo information. This data is usually regularly purchased from ISP’s (AT&T, Comcast, Virgin Media etc).
Some companies produce their own Geo data or make hybrid data sets. For example it’s common 3rd party knowledge that Yahoo fuse IP information with the user’s profile information. If the user has declared that they are in New York via their Email settings, then this may take precedence over IP data. Makes sense if you are on holiday in Spain as you don’t want to see local ads just because you are temporarily abroad.
Why is it inaccurate?
Inaccuracy usually stems from either guessing or dependencies. On the guessing side of things, IP addresses are dished out in big chunks (ranges). Since the entire range will be given to 1 provider it’s safe to assume that the entire range will be in the same country. Assumption being the mother of all…
The dependencies side is more important. Firstly lets assume that IP address information can change. If IP address 220.127.116.11 changes from being in Denver, Colorado to London, UK then how and when does you end product get informed? First the data has to get from the ISP to the Geo/IP provider, let’s say this takes 2 weeks. Secondly the data has to get from the Geo/IP provider to the end product. If the end product might update it’s database once a month, regular enough. This means that for up to 6 weeks the IP address would show as Denver when in fact it’s London. With over 4.3 billion IP addresses, 99% accuracy means 43 million IP addresses have inaccurate Geo/IP data at any time. Coincidentally that’s almost the entire population of Spain (or the population of California plus New York City, whichever statistic floats your boat).
Why Do IP Addresses change?
This is lovely and simple. There simply aren’t enough to go around.
IPv4 is limited to 43 million addresses. This is being replaced with IPv6 which is limited to 340 undecillion (yes that’s 340 with 36 0’s at the end!). This update will be complete at some point in the future, but changing the entire architecture of the Internet isn’t easy.
With a limited pool of IP addresses your Internet provider will only have a limited supply. To make sure they only use what they need, IP addresses will often “Dynamic”. This means each time your router reconnects to the Internet (after reboot for example) you may be given a new IP address. Many ISP’s offer “Static” IP addresses for extra cost, meaning that your regular cheapo Internet package will have a dynamic IP as standard.
IP addresses are a commodity and they are reused and reallocated when required. The updated information takes a while to reach the end products, and not all products get their information from the same place. When you compare 2 “like for like” reports, they probably aren’t “like for like”. If you delivered 10 impressions to Ghana when you targeted US IP’s, they were probably served to a little old lady in Texas… or maybe not! The data is good enough (statistically speaking) and will only get better as time goes on.
IPv6 will help solve everything (including but not limited to fixing the global economy)…. unless you subscribe to the theories that spam-bot’s will reap the dark rewards of a seemingly never ending supply of fresh and untainted IP addresses. But that’s whole other story!