When websites try to determine a user’s geographic location, this is usually done via the user’s IP address. In theory an IP address can be related to a physical location. This is how Google determines a user’s location in Analytics and Adwords – and even search – and Facebook in Insights.
It seems the accepted wisdom is that while IP address geolocation is not perfect, it’s mostly accurate. Estimates reach from 60% accuracy all the way up to 95% accurate. The problem is that the actual accuracy of IP geotargeting is nowhere near that.
On a country level, yes, IP addresses can usually be attributed to a specific nation. A Belgian IP address nearly always indicates a Belgian user. On a country level, the accuracy percentage of IP address geolocation does indeed tend to be upwards of 90%.
However, it’s when you get to regional IP targeting that all semblance of accuracy goes straight out the window, and it becomes nothing less than a total crapshoot. In fact, firing buckshot at a big national map will probably yield more accurate results than the geographical data you’ll get from any web analytics package.
I will give you a specific real world example that demonstrates the problem perfectly.
Northern Ireland is part of the United Kingdom (the other countries in the UK being England, Scotland, and Wales). For companies active in Northern Ireland, geographical targeting is very important. After all there is a small sea separating NI from the rest of the UK, so a business operating in NI cannot easily deliver services to consumers on the mainland UK.
We recently launched a Facebook page for a Northern Ireland business. This business is active exclusively in NI, and we sent out an emailer to their existing customers – all of which live and work in Northern Ireland, without exception. That emailer resulted in nearly 300 likes for the Facebook page.
However, when looking at the Facebook Insights for that page, specifically the list of cities the page’s fans are supposedly from, strange things ensue.
According to Facebook, the majority of the page’s fans come from London, followed by Manchester. Apparently only 34 of the page’s fans are actually from Northern Ireland, with the rest coming from either England or Scotland.
This is, of course, total bullshit. The problem is that it’s not presented as a ‘best guess’ – and a dreadfully inaccurate best guess at that – but as simple fact. The help text under the question mark reads: “Aggregated Facebook location data, sorted by city, about the people who like your Page.” Presenting such wildly inaccurate guesses as facts can lead to all kinds of confusion and misinformed decisions on the client’s side.
In Google Analytics things are a little bit better. The visitor location report for the client’s website at least shows that the most popular city is Belfast, and Londonderry is in the top 10. However I find it hard to believe that so many of the site’s visitors are from Glasgow and Edinburgh. The company doesn’t operate in Scotland at all, and its brand name would only draw blank stares over there.
The root of the problem is, of course, that these cities are where major UK ISPs have their hubs. So when users go online via these ISPs, their IP addresses will be associated with the geographic location of those hubs, which in the UK means primarily London, Manchester, Glasgow, Edinburgh, and a host of other cities – depending on where the ISP has capacity at any given moment.
This makes accurate regional geotargeting – or drawing any kind of conclusion at all from the regional geographic data provided by these tools – entirely impossible.
It’s not all doom & gloom. Geotargeting for mobile phones, for example, is exceptionally accurate. That is because it doesn’t rely on the phone’s IP address, but on the locational data derived from its GPS receiver or the nearest mobile signal antennas.
As laptops are increasingly supplied with built-in GPS chips, this type of nearly flawless geographical reporting can be done much more accurately than is currently the case. And on Facebook nearly all users fill in their profiles with the locations of their home and work.
But Facebook doesn’t use this data, instead choosing IP addresses to provide geographical information for its Insights reports. And as long as analytics tools like Insights and Google Analytics rely on IP addresses first for their locations reports, and fail to emphasise the inherent lack of accuracy of these reports, things will remain very murky indeed.