How Accurate is Google’s Geo-targeting?
Fall is officially here, and businesses are starting to ramp up their digital advertising campaigns in hopes of catching the eyes of thousands of individuals as the holiday season nears. Google has led us to believe that targeting users in all areas of the world has never been easier (as long as we use AdWords,) but is it as easy as they say it is?
A recent article from Search Engine Land, intrigued us here at Rocket55 and made us want to dig a little deeper to find out exactly how accurate Google’s location Geo-targeting is. This article couldn’t have come out at a better time, considering Moz just released the “2015 Local Search Ranking Factors,” and one of the updates included the note that “Google has gotten better at location detection.” After reading this article from Search Engine Land, it was amazing to see statistical data relating to how inaccurate Google’s location detection actually is.
One of the most shocking parts of the article was that, on average, there was an error of 334 miles in the US when using mobile devices. How they came to calculate this amount of error was by asking people to log in to Google Analytics, and then using the Real Time reporting tool to monitor their own visit. They made sure that their respondents regarding mobile were not connected to WiFi, but instead using their internal geolocation function. This result should not only surprise most businesses, but frighten them as well. This means that if you are targeting people in New York City looking for keywords related to your product, there is a significant chance that some mobile device users might be considered by Google to be “outside of the target location,” due to the fact that New York City is only 301 square miles.
The article went on to explain that within their testing parameters, 55% of users did not experience any location error, but when they moved the plus/minus to 25 square miles the percentage without error only increased to 67%. This left between 33 and 45 percent of the testing group within the false positive/negative group. (False positive meaning: someone Google identifies as being in the target location, when they are not. False negative meaning: someone who is in the target location, but Google identifies as being elsewhere.)
As was mentioned earlier, this controversial article intrigued the Rocket55 team, considering we use Geo-tagging for many of our clients. This topic was absolutely something to dive into, and find out if there might be potential prospects who are not being shown our ads due to a location error on their mobile device. (Also, according to this article, Google’s average error for desktop users in the US is 117 miles. That’s a significantly smaller radius, but still larger than Queens, NY.)
When this article surfaced, our first thought was research. We combed the Internet, from Moz to Search Engine Watch, to just googling it! Finally, after not finding any relevant articles related to this, we looked at Google’s process to see if this problem might be explained through the Google Developers site. We saw that Google talked about Geo-targeting accuracy and mentioned that, “Some browsers use IP addresses to detect a user’s location,” and then went on to explain how a user’s IP address – at best – provides a rough estimate of location. This could easily explain a 25-50-mile radius of error; however, 334 miles? It just didn’t compute, so, back to researching!
After investigating countless articles and forums, dating all from 2010-2013, it seemed as though this issue used to be a common occurrence. However, the lack of blog posts and article forums in the recent year would suggest that this error had been fixed. So, we kept digging and started testing. Actually, we came up with a few ways for YOU to start testing!
HERE ARE A TWO WAYS YOU CAN TEST THIS CONTROVERSIAL ISSUE:
1. If you want to test this dispute, you should increase an existing campaign’s target radius by 20-25%. This will ensure that users from the cities surrounding the target area will be exposed to your advertisement. Then, check with the new data to see if there is any increase in conversions by your target City/Zip Code. If there is an increase, you know that the geo-targeting error is happening to your campaigns, and you should adjust your target radius to a reasonable area around the city so you aren’t getting wasted clicks.
We would recommend leaving this test running for one-to-two weeks to collect measurable data that will ensure you are receiving as many views and clicks by your target population as possible. We want to be clear that targeting outside of your intended area is only for testing purposes, so once you receive the information you are looking for, we would advise that you return to your original settings in order to avoid wasted clicks and wasted resources. Remember, it’s not the quantity of clicks you receive that’s important; it’s the quality.
2. Another way to find out if any users are being missed within your defined target area would be to set up a specific landing page geared towards users within your target location. From there you can set up an AdWords campaign that directs users who click to that location-specific landing page. Then, using Google Analytics, you can compare the locations of the users that landed on this new landing page to the locations of the users being shown your ads in AdWords, and make adjustments where they are needed. (*One large caveat to this testing would be the (not set) location that Google attributes to a significant portion of your data. So, while you won’t be able to see all of the data, you may be able to see if there are non-targeted locations seeing your ads.)
Even though we haven’t committed to entirely embrace this new information from Search Engine Land’s article, we thought it was interesting enough to test and see for ourselves if this is a legitimate concern. We encourage you to look at your statistics and do the same, because you never know when generally accepted information might not be as solid as you thought. Good luck, and happy marketing!