In December 2014 Google enabled measurement of store visits, and a huge number of advertisers took advantage of it. And it makes sense, because users often search for a product online and then go to store/hotel/dealership or somewhere else and buy it. This stuff happens all the time, and Google puts a lot of effort to help us measure it.
As “store visits” become a new type of conversion used for AdWords account optimization, I spent some time looking under the hood of the technology that drives it. It’s pretty cool stuff, and it should revolutionize the way we all measure offline behavior.
Google has mapping technology that understands not only advertiser’s store location, but also its shape! The borders (and coordinates) of millions of buildings worldwide have been mapped very precisely which gives Google really strong starting point for understanding location.
Also, Google knows which stores are receiving visits by tracking Wi-Fi, cell tower and GPS location signals.
Google also takes advantage of Wi-Fi signal strength in many stores. They can measure signals to differentiate between visits to the store and visits to the store right next door. Compared to GPS, which doesn’t work well indoors, Wi-Fi-based location does a significantly better job, including multi-storied buildings.
So Google has loads of active reporters of location history and some really great maps, but that’s not the only set of data they use. There’s a survey panel of over ten million users that Google used to verify the accuracy of store visits data.
Surveys are used to calibrate algorithms and inform Google data modeling on location. It gets store visit estimates to a high level of accuracy.
If Google assumes there was a visit and then the panel confirms that visit, they certainly know the data checks out. If Google thinks there was a visit, but the panel data says otherwise, the model is going to be updated with that discrepancy.
Being really close to a store doesn’t mean someone really was in that store. So being near a store doesn’t automatically counts as a visit. There are additional considerations.
Thirty seconds visit isn’t the same thing as a ten minute visit. Really short store visit could mean that “shopper” just passed through a store. Also, too much time spent in store doesn’t mean someone converted. For example, employees who spend time at stores in long, discernible patterns aren’t counted as store visitors.
Store visits are estimates based on aggregated, anonymized data from a very large sample set of users that have turned on “Location History”. This data is then extrapolated to represent the broader population and only reported if it reaches a strict, highly conservative confidence level.
But, how can we, as advertisers, know that we can trust this estimates? Google use surveys to inform and validate the information as I said above, but the meat of the estimates is lots and lots of data.
If they aren’t confident they won’t show anything at all. So if you see any store visits, you can be sure that your ads are bringing people to your locations.
If you start including store visits as a part of your total number of conversions, you’ll get a much clearer view of the impact your marketing is driving (and has been driving all along).
