A Short Note On Google Location Tracking
Users of computers and smartphones are broadly aware that their location is being tracked by the providers of their operating system and apps (Apple represents about 40% of US phone share and 11% globally, with Google’s Android making up the balance). However, in my conversations with people they are often surprised when I tell them how much location data is recorded.
Both Apple and Google appear to track you with the same amount of detail, however Google gives you much more access to your location data. Google’s Timeline even shows you the photos you took using your phone at the location it recorded.
Even when you are not using your phone, location history may be recorded by apps you are using (e.g., on a laptop or tablet). So even Apple consumers who use Google apps like Chrome and Google Maps are likely to have some Google location history. A broad array of marketing firms also attempt to use details like your ip address and operating system to track you and infer your location.
Both Apple and Google provide ways to turn off the location tracking they do and to delete your location history. Some users, however, are likely to find access to their location history helpful for things like remembering details of past trips, or documenting where they were. In addition to time stamps, latitude and longitude coordinates, Google location data attempts to infer your activities, altitude, heading, and speed.
How Many Times A Day Are You Being Tracked?
I use an Android phone and regularly use Google apps like Chrome and Google Maps. I power my phone off at night. With that use pattern, my location was recorded 178,808 times in 2016. The average number of times my location was recorded per day was 495 times, or roughly every 2 minutes. On days when I used Google Maps for directions while driving, the number of location recordings exceeded 1,000.
Table 1: Number of Google GPS Coordinate Timestamps Recorded Per Day
Of the 366 days in 2016, there were only 4 days that Google did not have location history for me: the four days I was backpacking in an area with no cell signals. Interestingly, there was location history for the period I was travelling in a foreign country and had my phone in “airplane mode.” Because I accessed Chrome and Google Maps overseas using wireless, Google was able to accurately record my location (typically with 20 to 50 timestamps per day).
How Accurate Is The Location Data?
Accuracy is based on how Google tracked you. If it was by GPS, it is accurate to within about 65 feet. If it was by wireless router location, it is typically accurate within 300 feet. Cell tower derived locations are only accurate within about 3,000 feet. For each location timestamp, Google provides you with an accuracy estimate. Table 2 shows summary statistics on the accuracy of observations made of my location in 2016. Seventy five percent of the observations were accurate within 50 meters (approximately 160 feet).
Table 2: Accuracy Of Location Observations In 2016
One way to analyze the accuracy of the location data is to look at the last location recorded each day (i.e., where you spent the night) and “reverse geocode” the recorded latitude and longitude to lookup the street address. I found that the addresses were generally correct, but also often off by 4 or so house numbers or located on an adjacent street.
Tracking Your Activities
In addition to location, Google tries to guess what you were doing at the time they recorded your location using acceleration and motion sensors in your phone. They provide a “confidence” estimate for their guess. I extracted the guess they had the most confidence in for each time stamp. Fully 63% of the 178,808 time stamps they recorded were classified as “unknown” or no guess was made. By comparing their guesses to my journal, it is also apparent their guesses are often confused by activities at similar speeds. The large number of missing data and accuracy issues means this data item is not particularly informative.
Table 3: Activities Recorded In 2016 
Because the activity profiling is spotty and not particularly accurate, you can’t create Fitbit-type profiles of your exercise. However you can calculate from your GPS coordinates how much distance you covered per day. But there are several challenges in doing so. Because of the location accuracy issue, you can get slightly different latitude and longitude readings over time even if you are in the same place. This will make it look like you are moving, when in fact you were not. I restricted the data to those readings that were accurate within 100 meters. I also rounded the latitude and longitudes to two decimal points, which smooths out small oscillations in position. With those modifications to the data, Table 4 shows my median distance covered in 2016 was approximately 22 miles per day. It strikes me as a little high, but plausible. When I spot checked some distances against known long car trips, they were accurate. Distances greater than 1,000 miles were days on which I took flights. Thus even when your phone is off, your location is being noted if you check your flight progress on a tablet with Chrome using the airplane’s WIFI.
Table 4: Distance Traveled Per Day In 2016
Altitude, Velocity and Heading Profile
As with location, Google provides a “vertical accuracy” estimate for the altitude data it records along with your location. The median accuracy is about ∓ 20 feet. The biggest constraint in using this data is that altitude with a corresponding accuracy is only recorded about 1% of the time (based on this data set).
Table 5: Altitude Observations In 2016
Velocity and heading are also rarely recorded (my median heading last year was South!) and thus not particularly informative.
The significant amount of data Google collects about you is how they are able to make recommendations, estimate your commute time, and, of course, direct targeted advertising at you. If you don’t like the idea of being tracked in this way, you can turn off the location component of Google’s or Apple’s tracking and delete your location history. Of course other entities are effectively tracking you (from phone records, credit card transactions, etc.). Personally, I find location history helpful to remember details (where did we stop for lunch on our trip last year? etc.) and to document where I was.
Transparent and reproducible: All of the labeled Tables can be generated using the free, publicly-available R program and the R code in “Location.R” available on github to analyze your own Google location data obtainable from the links in the article. The inspiration for my R code was a blog by Emeline Liu.
 You can download your location history by clicking the “gear” icon on your Google Timeline page. It is in JSON format, thus some kind of programming -- like the R code linked above -- is required to analyze it. I found that my history went back to 2011.
 In contrast, my wife uses Apple products. Her location was only recorded by Google 418 times over just 59 days in 2016 (averaging about 7 recordings per day for those days). While Apple was recording her location in detail, Apple does not provide a way of downloading one’s location history.
 Their tilting category means “The device angle relative to gravity changed significantly” -- not enormously useful to me.