I'm definitely glad for this journalism, but find myself a bit surprised that a 515 page dossier's spiciest bit is apparently the metric referenced in the title (likelihood of customer attrition == 0.)
> While many of the data points in my McDonald’s report, when viewed individually, seem innocuous, the accumulation of information over years creates a creepily accurate picture of my habits.
They couldn't even really knock what data was being tracked. Like the app wasn't grabbing GPS or microphone sample just all the interaction data, things you would expect them to track but it creates an absurd corpus of personal information.
The question is what do they do with this information? Presumably the goal will be to get you to spend more on McD's and/or spend less on getting you to do that but it's not clear how will the dossier could help with that.
Writ large you can look at data across all people and spot trends (upward or downward), if you have a bunch of die hard customers that suddenly change behavior that's a warning sign. or you can see the effect of larger ad campaigns. If Target could figure out you might be pregnant from buying some hand lotion I have no doubt a good analyst with this data set across millions of users could extract a wealth of information. It probably also helps them judge revenue, if they know they can expect 2.16 visits over the next 6 weeks from 10,000 people and about how much they're going to spend you can extrapolate your future earnings with more precision than ever before.
At the micro level it's used to manipulate behavior, knowing your favorite items to give you discounts or to ping you when you might be most vulnerable. Maybe you have data that shows a customer spending more on particular days amid fairly random purchases, you might be able to construe those days are after payday and the customer is more confident purchasing larger ticket items. you might then target coupons for the days they might be getting paid.
Or maybe you see a trend where the person is clearly buying for 2, or 3 people and that changes. or suddenly they're getting happy meals implying they may have a kid now and you can send ads for the latest Disney toy.
Gather enough boring data over long enough and you can learn a great deal about someone.
I'm definitely glad for this journalism, but find myself a bit surprised that a 515 page dossier's spiciest bit is apparently the metric referenced in the title (likelihood of customer attrition == 0.)
Probably because individually it's rather banal:
> While many of the data points in my McDonald’s report, when viewed individually, seem innocuous, the accumulation of information over years creates a creepily accurate picture of my habits.
They couldn't even really knock what data was being tracked. Like the app wasn't grabbing GPS or microphone sample just all the interaction data, things you would expect them to track but it creates an absurd corpus of personal information.
The question is what do they do with this information? Presumably the goal will be to get you to spend more on McD's and/or spend less on getting you to do that but it's not clear how will the dossier could help with that.
Lots of things,
Writ large you can look at data across all people and spot trends (upward or downward), if you have a bunch of die hard customers that suddenly change behavior that's a warning sign. or you can see the effect of larger ad campaigns. If Target could figure out you might be pregnant from buying some hand lotion I have no doubt a good analyst with this data set across millions of users could extract a wealth of information. It probably also helps them judge revenue, if they know they can expect 2.16 visits over the next 6 weeks from 10,000 people and about how much they're going to spend you can extrapolate your future earnings with more precision than ever before.
At the micro level it's used to manipulate behavior, knowing your favorite items to give you discounts or to ping you when you might be most vulnerable. Maybe you have data that shows a customer spending more on particular days amid fairly random purchases, you might be able to construe those days are after payday and the customer is more confident purchasing larger ticket items. you might then target coupons for the days they might be getting paid.
Or maybe you see a trend where the person is clearly buying for 2, or 3 people and that changes. or suddenly they're getting happy meals implying they may have a kid now and you can send ads for the latest Disney toy.
Gather enough boring data over long enough and you can learn a great deal about someone.