Is GEO-IP reliable?

As soon as you suggest Geo-IP as a ‘solution’ for identifying the whereabouts of a user (browser), the invariable question arises: does it work, is it reliable?

As background, Geo-IP (often referred to as Geo-Targeting) allows a website or application to (attempt) to determine the geographical (physical) location of a user in order to target them with relevant and usually localised information. For instance, if I can be identified as being in North Sydney, the website will default my local store to North Sydney; if I am arriving from Queensland, default the homepage picture to something from Brisbane, and not Perth.

There is no completely reliable Geo-IP solution. There are just too many variables and nuances in the Internet which make that impossible. Indeed, some users are deliberately out there to trick Geo-IP services into thinking they’re somewhere they’re not. (Long story, not for this blog).

Any so the decision to employ Geo-IP comes down to the granularity of targeting you require, the amount of money you want to pay and the benefit Geo-IP will provide to the bulk of users it does correctly identify.

Let me explain.

Granularity of Targeting

Some Geo-IP services in the States claim to be able to very accurately target not only streets, though users based on likely voting intention (blue versus red suburb), age and so forth.

Indeed, you can bet that Google, combining all of its technologies and databases, is remarkably accurate when it needs to be.

Unfortunately, in a country as small as Australia, there has been less incentive for Geo-IP providers to invest in such accurate data and so street and frankly, suburb targeting is simply not accurate or even available in some instances. My general observation is that the closest we can hope for in Australia, is targeting of our local telephone exchange, and even that can be tenuous. Many corporate networks are configured such that their ‘local exchange’ as it were, is in another city or even country. Many of our US clients connect to the Internet via Fort Worth, Texas. (Why they do this is a different discussion).

Most Geo-IP services will allow you to target by suburb, region, state and country. My experience in Australia is that while it doesn’t hurt to try and target by suburb, regions (e.g. Sydney) are a better bet if you are happy serving only City or State-based content.

What to Pay

There are two costs when it comes to implementing Geo-IP.

The first is the cost of the Geo-IP service itself.

There are high-end providers such as Digital Element (www.digital-element.net) that charge a minimum of say US$500.00/month and up to US$10k/month for high-traffic, global websites.

There are entry-level solutions such as IP2Location (www.ip2location.com) that charge an annual, US$199.00 for some of their packages.

Some providers charge per-lookup (e.g. $0.001 every time they identify a user) and certainly, many providers will limit the number of lookups in their packages, in order to force you to buy a more expensive solution in order to cater for your audience. (Obviously, you need to know how much traffic your website/application generates before purchasing).

Some providers will host their solution for you, while some providers require you to download and install a database of IP addresses onto your own server, with database updates provided with some frequency as IP addresses move and change.

(A small tip: to reduce the number of lookups you make on users, and therefore the cost of the solution, cookie your users with their location on your first lookup of them so that you do not need to look them up again.)

When it comes to selecting a solution, you get what you pay for. Depending on how important accuracy is will determine at which end of the market you invest: cheaper solutions are not unreliable, they’re just not as reliable (accurate) as premium services.

The second cost of implementing Geo-IP is the implementation cost itself.

This is effectively the cost of the web developer implementing the technology into your website/application.

The actual implementation should not be complex or expensive. Identifying the location of a user is pretty straightforward, and very fast.

Where it can become tricky however, is what the website does with this data. If you are targeting location-specific content to users, the content needs to be structured in such a way that the website can do this. If you are implementing Geo-IP into a new website, this should be no issue as long as it is planned.

Retrofitting Geo-IP into an existing website might be harder however and before doing so, have the web developer estimate for the work. 

How accurate is accurate

Being identified as being somewhere you’re not, is frustrating.

Though how often does it actually happen?

For most of us, most of the time, Geo-IP is pretty cool. The website knows that I am somewhere in Sydney so you’re showing me the 02 number, and not the 07. What more can you ask for?

Clients often cite to me those funny, MSN Messenger-style chat popups on dodgy websites, inviting you to chat to a lonely 18-year old in Croydon Park: except, you’re in Avalon, 40 miles away. Surely Geo-IP is unreliable if I have to drive halfway across town to meet someone?

This is not a good assessment of how reliable Geo-IP can be, and is instead, a great example of how Geo-IP can be used very effectively.

If you recall, you essentially pay every time you look up the location of a user. The more targeting and accurate the lookup, the more you pay. For an adult service targeting millions of these chat-ads, the costs of targeting based on postcode would be considerable, and unnecessary.

These services likely target that you are in the greater Sydney area, and then either hit you with a random Sydney suburb to make it seem realistic, or the most central suburb in Sydney in order to attract they greatest catchment of possible users.

Any dingbat that thinks they’re about to chat with a lonely 18-year old is also probably prepared to drive 50 kilometres to meet said lonely girl. After all, the ads keep running and so they must be working.

They don’t need to be pinpoint accurate at all.

As long as you offer your users simple and clear mechanisms to change (and save) their location in the event that they are incorrectly targeting, most users will thank you for saving them the time of identifying their location.

Upsides

Unless pinpoint targeting is a non-negotiable for your website (in which case, you probably need to require that users personally set their location), Geo-targeting has only upsides in my opinion.

There is a direct correlation between more targeted, relevant, localised content, and conversion. Show the user a local number, and they’re more likely to ring. Show them the photo of their local sales agent, and they’re more likely to call.

The technology is by no means prohibitive to implement, and it looks great. Well implemented and executed, it is way beyond being a novelty and really can have commercial, revenue upside.

Just don’t be disappointed when you drive all night to Croydon Park.