Earlier this year, I was trading anecdotes with a couple of fellow marketing professionals who were struggling to make sense of the geolocation settings within Google AdWords.
Underlying everything we assume about geo-targeting is Google’s ability to correctly locate the user. If we are looking to serve ads to people in New York City, then (obviously) we need to target people who Google identifies as being in New York City. Equally, we need to be sure that we are not targeting people who are not in New York City.
Whilst this second statement may seem redundant, it actually isn’t — because, as we have discovered, there are two errors that Google seems to be making. The first is the false positive, the person that Google identifies as being in NYC, when in fact they are not. The second is the false negative, the person who is in NYC, but whom Google has identified as being located somewhere else.
To test Google’s geolocation abilities, we asked people around the world to check to see how accurate Google was at locating them.
We asked participants to log in to Google Analytics and then use the Real Time reporting tool to monitor their own visit — from both a desktop and a mobile device. With the mobile device, we asked them to ensure that they were not connected via WiFi , but rather using their mobile device’s internal geolocation function.
In total, 149 people responded to the survey, about half of whom were from the United States. While the sample size might not be statistically significant enough to draw broad conclusions, the results were nevertheless interesting.
The main takeaway is that Google seems to be much better at accurately locating a desktop device (d) than a mobile device (m).
The graphic below depicts the average error in miles for all respondents:
There were a couple of outliers from countries where we might expect the geolocation accuracy to be less than perfect. So we looked at the numbers for the US:
And for the UK:
Before those of you in the UK start preening yourselves for the better numbers, we should remember the relative sizes of these two countries.
In almost all cases, Google was far more accurate with desktops than with mobile devices, though the numbers for Europe were slightly more respectable.
There were some individuals who discovered that Google had them happily located more than 2,500 miles from their actual location. One or two users were identified as being in a completely different country, with one user on the Spanish island of Ibiza being confidently located by Google in Rome and another in Pakistan being located a few thousand miles away in the United Arab Emirates.
So what are the implications of this for advertisers?
For one, we have to reconsider the geo-targeting options that Google is giving us. Currently, once we select our geo-target, we can choose from three options:
If you are running a local campaign, which is when geolocation accuracy is going to be most important, the second option is the one you would likely choose rather than the first (much looser) default option.
In our study, only 55 percent of users were correctly located with zero error. Even if we allow for location error of 25 miles (a reasonable radius for most local businesses and Google AdWords Express), this number rises to only 67 percent.
That means that we are looking at a false positive/negative rate of anywhere between 33 percent and 45 percent.
Let’s now consider the implications of these two types of errors.
With the false positive, Google is serving your local ad to a searcher who is outside of your target area. If your ad simply offers your product or service without specifying your location, the searcher may well click on your ad only to discover that you are too far away — a wasted click.
If your ad copy indicates your location, you may discourage the click. This means you will save on the budget but reduce the click-through rate (which can, in time, impact Quality Score and ad rank).
If, on the other hand, the person is a false negative, this means your ad will not be served to someone within your target area despite the fact that they are specifically searching for your product or service — a missed prospect.
There is little we can do to reach a person in our target area that Google believes to be outside of it.
As Google encourages advertisers to look increasingly toward mobile advertising, the news that geo-targeting is actually worse on mobile devices than desktops is worrisome. Of course, where a person is connected to a WiFi network, their mobile behaves as if it were a desktop, taking the same location as would a desktop computer; however, as more and more people are searching on the go, the question as to precisely where they are going is one that becomes increasingly important. They may well be on the go but a few hundred miles from where Google believes them to be.
One final thought: If you are ever falsely accused of a crime — don’t ever call on Google to provide you with an alibi!