[RosanaAnaDana]

Total aside, but kinda relevant.

I lost my phone like a dufus about a two weeks ago. Battery died and I had no idea where it was. When I pulled the my google location history, it was too coarse to tell me anything other than 'at your house'. However, I was able to pull the raw data from google and post process it by time stamp into a series of rasters that were fine enough for me to see that the phone was definitely in the bedroom/ bathroom area. After processing my data, finding my phone took all of a minute.

[mehrdadn]

This is interesting. By the raw data you mean from Google Takeout, or something else? Did you have to filter/smooth the data points?

[machinesmachine]

Also interested. Could you describe the process?

[RosanaAnaDana]

Sure. I pulled my data in the standard takeout.google.com/ process. The result is a json (? iirc). I parsed the json into x, y, and rasterized it using gdal.

https://imgur.com/a/ZhyTsDV

See I wasn't sure if I lost it in the couch in the living room, the office, or the bedroom. This got me within a few feet.

[crazygringo]

That's amazing, so cool.

Is that something Google's not doing with the data, but could? (E.g. they don't because their maps don't show most houses precisely enough or whatever, so it wouldn't be useful?)

Or is it relying on the fact that you are triangulating or similar from the known exact position of your WiFi routers or similar down to the inch, and Google doesn't have any way of knowing that?

[Phlarp]

Or they do and have internally assessed that it would be too creepy to provide to users, but are happily using it to better target advertising.

[HALtheWise]

Or they have decided that it's too creepy to use at all, so they don't use it for targeted advertising. Seriously, why does everyone assume that companies are evilly cackling in volcano lairs? They know that violating user trust is really expensive and a bad idea.

By the way, I'm pretty sure I've seen that Google's advertising targeting is only allowed to use "neighborhood level" location, which is designed to be coarse enough to not allow specifying individual people.

[abakker]

Do you get toilet paper ads when browsing in the restroom?

[flattone]

Perhaps this level of location resolution is not stable enough at Google scale to present it? AKA: better to show reliable fuzzy information than unreliable precise information.

[dannyw]

Considering that google has a history of cloaking information via the UI (see purchase history hidden if you have G Suite, but still fully accessible via takeout), and that Google offers advertisers the ability to see if you have visited a particular store even in an indoor mall, I am sure google knows your location more precise than it reports.

[gricardo99]

This is pretty cool, but I'm still confused on what data you used.

By my data, do you mean data from Google Android Device Configuration Service?

If you're logged into Chrome or GSuite tools from desktop locations, I just wonder how useful the data from those other products would be, if it even has location data.

I'm downloading my data archive to check it out...

[RosanaAnaDana]

I think its https://takeout.google.com/ where I got the data. Then its just a json which is relatively easy to parse.

[icelancer]

Ha, this is an awesome use of the data. What a cool plot showing the "density" of pings.

[jdironman]

Probably just looked at the last data point depending on the type of data was available. If its longitude / latitude values then its going to be pretty precise, although I would also like to know where to pull more granular data.

[RosanaAnaDana]

The GPS receiver isnt' sufficiently accurate for one point to be useful, but in aggregate, it was accurate 'enough'.

So although the numerical precision is pretty high, the actual accuracy is pretty low, I think I pulled all the points from 2am-6pm in the final time stamp to figure out where the phone was.

[nerpderp82]

Envious, nay jealous of a room of that size.

[raxxorrax]

Probably slower than just keep searching. But way cooler.

[RosanaAnaDana]

Well, I looked for a week-ish, and was fap-all out of ideas where it was. I was at the point where I was looking in ridiculous locations (freezer, garage, washing machine, garden, tool shed).

This took about 20 minutes to hack together, and I found the phone in 10 minutes of looking in the 'right' spot. My phone had fallen behind the bedside stand and found a way to balance itself between the bed and the baseboards. The bed post is ~phone width, so even looking, you just couldnt see it.

That being said, I work in the geospatial sciences doing geospatial data processing, so 20 minutes for me may not be 20 minutes for some one else.

[raxxorrax]

> freezer

Yeah, I have these moments too. Could be a good disinfectant for keys, I guess.

[hugi]

Plot twist: That's actually the queen and she lost the phone in Windsor Palace.

[lawrenceyan]

Point clouds save the day!

[rmorey]

This rules, thanks for sharing!