How Facebook Knows When Its Ads Influence Your Offline Purchases | Hacker News Reader

How Facebook Knows When Its Ads Influence Your Offline Purchases

wired.com

52 pointscyphersanctus47 comments

...aaand facebook starts to buy news pieces to justify their ad network.

this is nothing more than a press-release for their atlas feature. A new low even for wired.

Odd. It seems unusual for a press release to include criticism from the EFF.

How would the article have to have been written for you to not think it was a press release?

I feel like PG's submarine piece has really made HN's view of PR very black and white. There are a lot more gradations, and likely this is a piece that was pitched by Facebook and actually was moderately investigated by Wired. The worst journalists are copy pasting press releases, the best are finding their own stories, but the vast majority lie somewhere in between. And I'm not convinced there is anything wrong with that.

Criticism? The EFF comments were practically praising Facebook:

  From the outside, says Adi Kamdar, an activist with the Electronic Frontier
  Foundation, it seems Facebook is “doing a good job” anonymizing personal
  info as it seeks to link online and offline behavior. Facebook is adamant
  that its system can’t be cracked—i.e. can’t be used to match names from
  Facebook accounts with behavior off of Facebook—and Kamdar says he has no
  reason to believe otherwise.

Some commentators in this discussion question why anyone would give their phone number to shops when they make a purchase. Yet millions of people give their mobile phone number (and name, date-of-birth and gender) to Google when they sign up for a Google account. And yet the degree of tracking that Google can undertake is an order of magnitude greater than what any individual store can track. Does Google use your mobile number solely for two-factor authentication and absolutely nothing else? Who knows? Google doesn't tell you.

When you can sign in on your tablet/mobile/desktop/TV/thermostat/fridge and who knows what else - how hard can it possibly be to stitch together your spending and consumption habits with all the other data collected about you? In fact, it wouldn't surprise me if those joined-up journeys have already been compiled. It's pretty obvious that the information companies collect about you goes far, far beyond creating a simple "advertising profile".

Companies like Google and Facebook (and others) have no self-restraint when it comes to tracking you and they don't even do it anonymously. How do they get away with it? Because we happily let them.

Sure. I think the difference is that Google hardly gives you a choice, other than not using their products. Not that this is a great situation. And it still doesn't answer the question why people would give away the same information voluntarily without a direct need.

Google uses Android phones GPS location to track people's store visits.

http://digiday.com/platforms/google-tracking/

http://www.datadrivenbusiness.com/google-quietly-testing-off...

If you turn it off, things like Google Now stop working.

Yep, im pretty sure just the mac address collection from wifi radios would be enough to track what store you have been in if you run an Android phone.

I have wifi and gps turned off except when i am lost.

> The trick is that the hash of a phone number captured on Facebook will look just like the hash of the same phone number captured in a brick and mortar store, so the two companies can match the numbers without actually trading them.

This just isn't possible. How many cents would it cost to brute-force hash all legal American phone numbers?

Why on earth would you have to brute force all numbers? You can do this optimally: just hash the numbers you see (or get requests for).

So that I can cold call someone who might be interested in my product, even though DataLogix only told me the hash of their phone number.

You've already hashed all numbers on your end, so it's trivial to figure this out (hash the hash?). Hopefully I'm not thinking about this wrong, but it does seem feasible.

This is correct. As a zero-knowledge proof it is, sadly, trivially broken. (Not just in a theoretical sense. As you mention there just isn't enough entropy going into the hash. It's like asking for your age and gender, but only providing a hash to some other party so they can verify that you are who you say you are, without your divulging what information you were just given. That doesn't work - you would leak everything, because 1-100 (age) M/F (gender) only has 200 possible values. 200 hashes later your counterparty knows what you were asking to check.)

But that doesn't mean this couldn't be done properly.

There are actual zero-knowledge proofs. I liked this primer. You should read it!

http://blog.cryptographyengineering.com/2014/11/zero-knowled...

It is 100% possible for there to exist absolutely zero-knowledge proof in many instances. (Such as the one in the article.)

So, it could be possible for example (I don't know an algorithm) to check whether a phone number you were given is in someone else's set of phone numbers - without either your learning what the other's set is, or the other learning what phone number you're testing.

If there is, the 'someone else' would still have to do some rate limiting. Otherwise, once again, I can just run this algorithm for all possible phone numbers.

No, no. Read my link :) The whole point is that the definition of 'zero-knowledge' is that (if you use a zero knowledge interactive protocol) you learn no further knowledge.

You are right that the hash-passing is not zero-knowledge. Rather than just verify whether it's in a set, you've accidentally revealed a number. So if you have a set of phone numbers and I have 1 to check if it's in that set, then checking hashes doesn't work - I would end up divulging my number if I used it.

Intersecting a set of observed identifiers against another set is an issue Moxie dealt with on the TextSecures blog[0] (for phone numbers at that). He evaluated some ZKP algorithms and basically concluded this problem is hard or impractical today. I imagine an ad/referral network has many of the same real time constraints and scalability issues.

[0] https://whispersystems.org/blog/contact-discovery/

thanks - this was interesting. Some of the numbers are interesting. So, for example, he says to have ten million clients updating daily, he would need to sustain 40 MB 116 times every second. . . that's like 37120 megabit so if you pay $10 per megabit monthly, that's $371,200 per month. It's pretty bad. On the other hand you do have ten million users.

so the numbers are off, but they're not five orders of magnitude off. If we're going to service ten million users, we might have something of a hosting budget for it.

So while I believe the author that the current trade-offs aren't goods, clever academic mathematics might help in the future. Cryptography itself came from there.

Many other apps hash phone numbers to "suggest" security and privacy as well, e.g. Secret. Other apps don't even bother to create the hashes and send the plain text phone numbers, e.g. WhatsApp.

I see false positives. So last week I decided to do my first project that will use a Raspberry Pi. I googled around to the pi site, amazon, newegg, (gasp) radio shack, microcenter, etc... I figured out what I wanted to get but didn't put anything in any carts. So then Pi ads start showing up on Facebook. IIRC a Radio Shack Pi ad too. Whatever. But now when I buy one, is Facebook going to claim some responsibility for the sale? Because they honestly don't deserve it. I'm going to buy one anyway and I have not seen an ad for a place I didn't look already. While I often give a real store the business based on "I can get it on the way home" the price difference in this case is nuts. So NewEgg it will be and Facebook ads aren't going to influence that.

In this case, correlation does not indicate causation.

causation is implied on a population through test and control. causation and correlation for an individual is almost always impossible.

So are they going to use statistical models to figure out how much to charge for ads? That seems like a hard sell.

"It has happened to you, and it has probably happened more than once."

Umm, no. Why on earth would I give my phone number or email address to random shops? This is crazy. I surely don't want merchants to contact me "about other stuff I might want to buy." This assumption sold as a certainty in this article is somewhat disturbing. Do others from the US feel the same way? (I'm from Germany. Different culture?)

I can only think of a couple of retailers that have asked me for an email address here in Canada (coincidentally both were US companies).

I do get asked for my postal code fairly often but that's hardly a personal detail.

Same here regarding the postal code, that's fairly common in Germany.

You postal code, name and the expiration date on your card is enough to positively ID you in many consumer databases, and by giving it to the retailer you're implicitly granting consent for its use for this purpose.

I am not a lawyer, but I'm pretty sure that in Canada, this would be in violation of the federal "Personal Information Protection and Electronic Documents Act", as well as many provincial laws.

https://www.priv.gc.ca/leg_c/leg_c_p_e.asp

"Use or disclose personal information only for the purpose for which it was collected, unless the individual consents, or the use or disclosure is authorized by the Act."

"Keep personal information only as long as necessary to satisfy the purposes."

If the client uses one of those "points" cards, yes, they are tracked, but otherwise they shouldn't. When a retailer asks for your postcode, it is mostly just for market study to identify where their clients live, but that's it (although I rarely get asked). You can also just answer "no thanks" to this question.

Store memberships for points and rebates on products are really common here in the US. When you sign up for them, they ask for your email, phone number, address, etc.

Safeway, a big supermarket chain, does this to an extreme. Many of their products are "50% off" when you enter your phone number of your membership on checkout. (It really just brings it down to a normal price).

Nothing prevents you from using a fake number, but as far as I can tell, most people give a real number.

It's certainly a different culture in Germany. Media and people are a lot more concerned collection of data.

I use a fake number and a fake name at Safeway, and they always address me by the fake "Mr. XXX" when I check out. Always a little weird.

"People voluntarily link things like phone numbers and email addresses to their Facebook accounts"

People voluntarily are idiots. Never connect a phone number to your FB account and use another mail address if you really need to have an account.

I agree with you, but sadly it's not that straightforward; It's entirely plausible (perhaps even likely) that your "facebook friends" have your number, and do something stupid^H^H^H^H^H^Hconvenient like give fb access to their address book (as is often requested, and I wouldn't be surprised if it happens by default with their mobile apps), Facebook are only one name match away from having your number and perhaps DoB and other personal info anyway.

>As Boland explains it, Facebook simply shows advertisers that a given number of people who saw an ad for a product also purchased that product.

Facebook doesn't share my personalized shopping data with advertisers. That's good.

But the data still exists. So the data the advertisers receive is only really anonymous until there's some sort of security breach or economic decision whereby it's suddenly more profitable to sell the personalized data. Either or Both of which are inevitable.

All in all, a moderately informative marketing piece.

Facebook doesn't share my personalized shopping data with advertisers. That's good.

The stuff about you personally that can be gleaned from market analytics will make you shit a brick.

And this is what private industry can do, not even getting into three-letter agencies.

> Facebook doesn't share my personalized shopping data with advertisers. That's good.

Facebook wants to keep the data inside their keep.

tldr; it doesn't, but offline vendors that collect your contact information they can anonymously check the proportion of purchasers which might [incidentally?] have been shown their ad on Facebook's network. Only a media buyer with a budget bigger than their brain would assume causality.

Couldn't Facebook easily demonstrate causality by running an RCT? Just randomly select half of a certain demographic to show the ad to, and then show that those users are more likely to make a purchase later.

Theoretically yes, but I think with standard Facebook ad units and regular consumer brands sold in stores they'd be pretty unhappy with the results...

(since probability of giving stores your contact details post purchase and susceptibility to marketing messages pre-purchase are probably positively correlated and Facebook is permanently running concurrent trials of different kinds and being very selective about what they share, any positive effect they did manage to identify should be viewed as an almost certain overstatement of the ads' actual effectiveness anyway)

Who gives their phone number to store clerks? "Sorry, it's unlisted."

It's s numbers game. If store clerks ask everyone, all the time, then enough people will answer for the practice to be useful.

As hard as it is to believe, I'm sure plenty of people give out their phone numbers when asked. I have personally witnessed many many people (including my wife) give out their number at checkout without hesitation. I guess the fact that this is continued practice means that it actually works.

This may vary by geography...my observations are from Canada.

Large department stores would frequently pitch a loyalty card that's then tied up to your profile, for a which a phone number is frequently a must.

I'd argue that any HN member (or even any 1/2 tech savvy consumer) isn't exactly their target market

Is anyone else frightened the increasing efficiency of capitalism? Since when does user-experience depend on best fulfilling my consumer desires? Shouldn't I try to soften this impulse?

And this is why I did not give my phone number to facebook and why I use a different email for everything.

It's unfortunate that keeping one's privacy requires a lot of knowledge and work.

> It's unfortunate that keeping one's privacy requires a lot of knowledge and work.

I wonder if the more vexing problem is that the market for privacy solutions seems largely unproven at scale. While people seem to value privacy when polled individually, they seem to vote the other way as an aggregate body.

Smart privacy products could take the knowledge and work out of it, but will the market support them?

Amazon product pages already have Facebook tracker embedded, so Facebook knows what you are looking at. No need to provide your phone number.

Is that visible via NoScript?

I'm using disconnect.me

Reply on news.ycombinator.com