Starting today, the first time you turn on Location History—which is off by default—your auto-delete option will be set to 18 months by default. Web & App Activity auto-delete will also default to 18 months for new accounts. This means your activity data will be automatically and continuously deleted after 18 months, rather than kept until you choose to delete it. You can always turn these settings off or change your auto-delete option.
If you’ve already had Location History and Web & App Activity turned on, we won’t be changing your settings. But we will actively remind you about the auto-delete controls through in-product notifications and emails, so you can choose the auto-delete setting that works for you.
[...]
[...] we're bringing this to YouTube, where auto-delete will be set to 36 months by default if you create a new account or turn on your YouTube History for the first time.
[...]
We’re also making it easier to access Incognito mode in our most popular apps, by long-pressing on your profile picture in Search, Maps and YouTube. It’s available today on the Google App for iOS, and coming soon to Android and other apps. We’re also working to make it possible to stay in Incognito mode across Google apps, like Maps and YouTube, and will have more to share soon.
I very much doubt they would risk their reputation by outright lying about this for fear of having that fact leak due to a whistleblower, especially since that would amount to fraud and the...
I very much doubt they would risk their reputation by outright lying about this for fear of having that fact leak due to a whistleblower, especially since that would amount to fraud and the executives responsible could actually go to prison for it. And I think it should be pretty obvious to everyone that they still intend to keep the derivative data, since this is a user data retention policy update, not a complete business model overhaul.
That's a problem with just about everything remote. It's pretty close to impossible to verify what's running on someone else's hardware. At some point you just have to trust. That said, Google...
That's a problem with just about everything remote. It's pretty close to impossible to verify what's running on someone else's hardware. At some point you just have to trust. That said, Google sure hasn't earned the trust and keeping the aggregated statistics while deleting the source data does sound like something they'd do.
By the time your data is 18 months old, it's useless. Heck, I'd bet that even after a week it's a lot less useful for ad targeting than it is if they can put it to immediate use.
By the time your data is 18 months old, it's useless. Heck, I'd bet that even after a week it's a lot less useful for ad targeting than it is if they can put it to immediate use.
Not really. Especially for location data. People typically commute to the same places, buy at the same shops, see the same friends. Forget an algorithm, even a person could identify, with...
Not really. Especially for location data. People typically commute to the same places, buy at the same shops, see the same friends.
Forget an algorithm, even a person could identify, with someone's location data from 3 years ago, where they work, their rough income level, political affiliation, social class, race, age, and marital status.
None of those things change very often, so it's very valuable.
With it not changing any time, it doesn't seem like it'd add any additional value over the stuff within 18 months. 18-month-old data on its own, sure, that's worth something. But I don't think it...
With it not changing any time, it doesn't seem like it'd add any additional value over the stuff within 18 months. 18-month-old data on its own, sure, that's worth something. But I don't think it is in the context of the 18 months of newer data still hanging around.
Very true. Although someone else made the point that just because the source data is deleted doesn't mean they can't still store the relationship between you and your university that was derived...
Very true. Although someone else made the point that just because the source data is deleted doesn't mean they can't still store the relationship between you and your university that was derived from that source data. Honestly kinda makes sense since it'd be incredibly hard to keep track of what derivations came from what source and to track what needs scrubbing when its source goes out of the 18-month scope.
Not an ad guy, but as a data analytics guy if I wanted to understand people's travel or vacation habits I would need way more than 18 months. People typically just take 1 or 2 trips a year. If I...
Not an ad guy, but as a data analytics guy if I wanted to understand people's travel or vacation habits I would need way more than 18 months. People typically just take 1 or 2 trips a year. If I want to understand patterns I'd need several years of data to understand what kinds of places people like to go. Like, are they basic or do they go off the beaten path? Do they like to go on a budget or blow a ton of cash? Do they like spa and resort type places or do they prefer modest accommodations?
Granted I could probably infer that from lots of other sources of data, particularly what kinds of feeds or reading material you subscribe to. But then that casts doubt on the value of tracking cookie-based analytics in the first place. If the very fact that you subscribe to Kinfolk and Monolith can tell me more about you than all the cookies in the world, what's the point of the cookie? (This is a bit self-serving though. I've been arguing for a long time that this sort of invasive tracking is really low value and doesn't really help with much in the way of insight generation).
Yes it can be useful, but on the other hand if you use data going further back it also risks being irrelevant because the world has changed. They got married, had kids, moved to a different state,...
Yes it can be useful, but on the other hand if you use data going further back it also risks being irrelevant because the world has changed. They got married, had kids, moved to a different state, got a better job, lost their job, took up a different sport, or just changed their vacation preferences. Or there was a pandemic.
No matter what you do, historical data only gets you so far. It will work best for predicting the habits of predictable people in stable times, but then again that's the easy case.
I think it's true in general that older history is less useful than newer history. It also costs more to keep, and they may have decided it's not worth it. Twelve years ago, Google made a similar...
I think it's true in general that older history is less useful than newer history. It also costs more to keep, and they may have decided it's not worth it.
Twelve years ago, Google made a similar announcement about how long to preserve search traffic and how to anonymize IP addresses. (For people who are not logged in.)
This is about the raw data - non-individual summary statistics might be kept for longer.
From the article:
[...]
[...]
How can we be sure the really delete it. They might delete the history, but not the statistics they got from it.
I very much doubt they would risk their reputation by outright lying about this for fear of having that fact leak due to a whistleblower, especially since that would amount to fraud and the executives responsible could actually go to prison for it. And I think it should be pretty obvious to everyone that they still intend to keep the derivative data, since this is a user data retention policy update, not a complete business model overhaul.
That's a problem with just about everything remote. It's pretty close to impossible to verify what's running on someone else's hardware. At some point you just have to trust. That said, Google sure hasn't earned the trust and keeping the aggregated statistics while deleting the source data does sound like something they'd do.
By the time your data is 18 months old, it's useless. Heck, I'd bet that even after a week it's a lot less useful for ad targeting than it is if they can put it to immediate use.
Not really. Especially for location data. People typically commute to the same places, buy at the same shops, see the same friends.
Forget an algorithm, even a person could identify, with someone's location data from 3 years ago, where they work, their rough income level, political affiliation, social class, race, age, and marital status.
None of those things change very often, so it's very valuable.
With it not changing any time, it doesn't seem like it'd add any additional value over the stuff within 18 months. 18-month-old data on its own, sure, that's worth something. But I don't think it is in the context of the 18 months of newer data still hanging around.
Old data can be useful, too, though. For instance, if it had the last few years of my location data, it would know where I went to University.
Very true. Although someone else made the point that just because the source data is deleted doesn't mean they can't still store the relationship between you and your university that was derived from that source data. Honestly kinda makes sense since it'd be incredibly hard to keep track of what derivations came from what source and to track what needs scrubbing when its source goes out of the 18-month scope.
Not an ad guy, but as a data analytics guy if I wanted to understand people's travel or vacation habits I would need way more than 18 months. People typically just take 1 or 2 trips a year. If I want to understand patterns I'd need several years of data to understand what kinds of places people like to go. Like, are they basic or do they go off the beaten path? Do they like to go on a budget or blow a ton of cash? Do they like spa and resort type places or do they prefer modest accommodations?
Granted I could probably infer that from lots of other sources of data, particularly what kinds of feeds or reading material you subscribe to. But then that casts doubt on the value of tracking cookie-based analytics in the first place. If the very fact that you subscribe to Kinfolk and Monolith can tell me more about you than all the cookies in the world, what's the point of the cookie? (This is a bit self-serving though. I've been arguing for a long time that this sort of invasive tracking is really low value and doesn't really help with much in the way of insight generation).
Yes it can be useful, but on the other hand if you use data going further back it also risks being irrelevant because the world has changed. They got married, had kids, moved to a different state, got a better job, lost their job, took up a different sport, or just changed their vacation preferences. Or there was a pandemic.
No matter what you do, historical data only gets you so far. It will work best for predicting the habits of predictable people in stable times, but then again that's the easy case.
I think it's true in general that older history is less useful than newer history. It also costs more to keep, and they may have decided it's not worth it.
Twelve years ago, Google made a similar announcement about how long to preserve search traffic and how to anonymize IP addresses. (For people who are not logged in.)
This is about the raw data - non-individual summary statistics might be kept for longer.