This doesn't seem to describe a Ponzi scheme. Money from later investors is not used to pay returns to earlier investors.
Contrariwise, if location data from later subscribers were used to provide solid information to earlier subscribers that would be an effective and useful service.
If there is collusion to both sell fake data and detect fake data that is indeed fraud, but it's a different kind of fraud.
What the article is getting at, is that the service makes promises based on the assumption that later subscribers will show up to provide enough data to satisfy the SLA they give to the early subscribers, before those early subscribers attempt to "call them on it" (by i.e. actually attempting to use the less-common data sets in production.) If the later subscribers don't come, the service will eventually be found useless for the early subscribers, and so they've paid in for nothing.
Thus, like in a (complete-knowledge) Ponzi scheme, the early subscribers have an incentive to get later subscribers to sign up for the service, so that it can be more useful for them.
But, unlike in a Ponzi scheme, the later subscribers don't have to wait for subscribers even later than them, before they can benefit. The later you subscribe, the more immediate value the system has to you.
You're doing a good job of retconning the article but here's what it actually says about Ponzi:
"There are ad tech companies that promise ad buyers they will find fake data. Those vendors will usually eradicate that data at a cost-per-thousand, as a data fee. Similar to what happened with viewability and ad fraud. I’m sure there are companies looking to solve this problem, but you’ve got to wonder: Why would I want to pay more to validate the data that I already paid for? If this happened in the financial industry, then people would’ve been locked up for it — it’s like Bernie Madoff’s Ponzi scheme. Companies that detect fraud would not need a reason to exist if the market didn’t pay for the fraud to take place."
The bit they explictly compare to Ponzi/Madoff is having to pay an extra fee to detect fakes in data you bought because the "market" is paying for the fraud to take place.
This is almost, but not quite, entirely unlike a Ponzi scheme.
Also, this is terrible journalism. It's almost impossible to tell which statements constitute the "confession" and which constitute editorial comment.
It's more like "trying to kickstart the network effect".
It's kind of like how dating services buy and sell portfolios of user accounts so that they can pre-seed their services with apparent potential matches, thus attracting real users who will (hopefully) become the real userbase.
The early arrivals are in an empty room. If nobody comes after them, they won't have much fun. If you arrive after a critical mass of people is already present, though, the value of the party is instant.
I guess the difference here is that people know how parties work in this respect; this business may not have been so clear.
I believe that's called "an investment" -- you put resources in, hoping to get them back again plus more. The possibility of losing your principle is called "risk".
It's about as much of a Ponzi scheme as, say, a day at the track.
If you paid for a service, but got an investment, that's not exactly on the up-and-up. If that money is then used to try to work towards what they've sold you as currently available, that's not exactly a Ponzi scheme, but it does share quite a few similarities.
IMO, equating it to betting at the track isn't very accurate. Most people that bet at the track know they are taking on risk of no return, and what they gain for that is increased payout. Where's the knowledge here that you might not get what you paid for? What's the benefit of taking on this risk? Is it still betting if you buy something at a store and the shopkeeper turns around and puts the money into a slot machine before you've been given what you paid for and before know what's going on?
The author's words were that the sale of fake data was "like a Ponzi scheme" in the sense that it is a crime.
The author makes no mention of any factual similarities.
The author is suggesting there are legal similarities: both are criminal fraud. He suggests that those committing fraud on investors are "locked up" and insinuates those committing fraud on advertisers are not.
The author were "[i]f this happened in the financial industry..."
If what happened? Sales of ad data? No. Fraud.
The author is not suggesting that deceiving purchasers of location data is factually similar to deceiving purchasers of investments.
The author is suggesting they are both fraud, i.e. legally similar. Both are intentional deception by a seller on which a purchaser relies. Fraud.
The difference the author sees is that those who commit fraud on investors are incarcerated and those who commit fraud on advertisers are not.
Maybe the author was just trying to cite a well-known, real world example of fraud (intentional deception by seller upon which buyer relies) where the perpetrator was incarcerated. In that case, the example he picked makes sense.
My problem is that the article is titled "Confessions of a location data exec: ‘It’s a Ponzi scheme’," so the question of whether it's a Ponzi scheme or not is at least relevant.
As for the analogy, it's possible this exec is just that bad at language, sure. "My darling, you are like a fine Stradivarius violin... it is possible to make you useless by lighting you on fire."
Oh, I missed the title. Maybe I am wrong then. But then what good are titles anyway? HN routinely changes them from their originals. Reminds me of patents. One cannot derive anything useful from the title alone.
I pay a lot of attention to stuff like this because it's something my customers ask for. Podcasters pay good money to hosting services for demographic data on listeners. There's no good way to get that data: podcasts are RSS feeds. You see IPs and user agents. How do they know who's a 30 year old male making $70,000? You an try to match IPs with advertising databases, but good luck getting any sort of real signal out of that. That data is all inferred anyway. Hell, I bet my cell phone has two dozen IPs over the course of an average day as I connect to various wifi networks and Fi switches carriers.
And yet, people build million dollar businesses on this stuff. Blows my mind.
Edit: And yes, sometimes people play podcasts via players on the web. This makes up a tiny fraction of overall listens. It's not statistically significant.
There's no good way to get that data: podcasts are RSS feeds. You see IPs and user agents. How do they know who's a 30 year old male making $70,000? You an try to match IPs with advertising databases, but good luck getting any sort of real signal out of that. That data is all inferred anyway. Hell, I bet my cell phone has two dozen IPs over the course of an average day as I connect to various wifi networks and Fi switches carriers.
As it damn well should be.
That said, it's only a matter of time before carriers start tracking and selling this data.
It's a lot more involved than that, but reputable companies are few and far between. They reason there's so much fraud is because it's hard to prove the truth and fake results look "good" so people go with shady vendors that just make their marketing team look better.
We do account-based marketing (advertise to specific companies you want to sell to) but we're upfront on the minimum size that companies need to be to have a internet footprint. There are other methods to find people working at a company but one thing we see all the time is companies claiming to let you reach the CFO or an exact job title. That is 100% fake and it would be far more effective to just send a letter to that person instead, but people just love to believe it works.
As an Location Data Exec I can say that so much of what he 'assumes' here is utter tripe.
There is no need to have the phone in hand with the application open in order to collect location data - we get billions of signals from first party data from background usage on both iOS and Android.
These SDK's require legitimate usage of GPS in order to deliver the functions that we provide - our permission requests are overt and informational - we let users know exactly what data they will be sending and how we use it (one of our products actually rewards users for doing this!).
True, Apple and Android are both cracking down on un-authorised usage and collection of this data, it just means you have to follow some more rules in order to collect this data. Either way it is good for user transparency.
It sounds like (for the most part) he is talking about bid stream data, which everyone in the industry knows is the sewerage of the location data industry. Sure you can get a few valuable insights from it - but should it be classed as location data? No way. Even with significant cleansing it is still mostly garbage. If that is what their entire business model is based upon then no wonder it feels like a ponzi scheme to him.
For the rest of the location world who are using quality data - we see things differently!
Not an exec but work in the geospatial data space.
Our company did an analysis of bid-stream data and came to the same conclusion that less than 10% of bid stream data is high quality.
Yep we know your company very well - we came to the same conclusions, even despite the heavy cleansing methods people purport to undertake it is still garbage!
> we let users know exactly what data they will be sending and how we use it
Many apps will just say “see privacy policy” or similar when requesting access. See [1] for examples. That is an approach I would consider dishonest as most users will not try to find and read through it. Does your method use a more explicit consent flow? I think that is a big problem in this space, so you could really differentiate yourselves if informing users in a truly explicit manner unlike others.
It is a huge problem in this space, we see that. We also believe that transparency and data democratisation is a huge draw for users.
I think everyone is aware nowadays that if a product is free, you, are in someway the product.
One of the things we wanted to achieve with our Reward product is that users not only know that we are collecting data but can actively monitor what data we are receiving from their device. Most importantly they can see a independently verifiable 1-1 relationship of rewards for each days worth of data sent.
It isn't much in the grand scheme of things but your data has value, you could argue that this data is payment for such "Free" services such as Facebook and Google - but transparency of this is important.
I really like what Brave and BAT is doing for turning the advertising paradigm around - part of me thinks that it is against human nature and it won't work but we literally have an entire generation of people who do things for "Likes" or "Upvotes" so maybe I am just wrong (I hope I am!)
The UUID (IDFA / AAID) is 'true north' for us, this allows us to join it to other data-sets, however there are several other methods of attribution we use.
We don't collect any PII data (in fact we actively avoid it!), we don't need to know 'who' a device is, only the location history of that device!
There are companies out there who will attribute a UUID to a physical email / telephone number, I think it is a bit of a grey area and with the current privacy landscape something I think that will soon end.
> they believe up to 80 percent or more of the lat-long data available there is fake. No one has stopped to think about where that data has come from and why a publisher would choose to sell it all to a vendor who is going to build a business on top of their data. What’s actually happening is these ad tech vendors are trying to pad out the limited data they already own with other data sets from competitive vendors or other unknown sources.
Sounds more like textbook fraud than a Ponzi scheme (though, honestly I did not read the whole article).
Not really fraud, phones have different levels of horizontal accuracy based off of privacy settings, GPS & cell signal strength, battery life, CPU load and that's before you get to the obfuscation that exchanges do. At Dstillery we built a geodata classifier to tell us what was and what wasn't good data. We throw out 60% to 75% a day as not useful for learning anything from. But we can be picky since we combine web and location data we aren't beholden to needing the unreasonable amounts of location data you need working with just location data. Location should be holistic part of the data, not the be-all, end-all.
It should also be noted that we have anti fraud tech baked into our system that fires before our geodata stuff runs. Fraud gets cleared out for being fraud not for being bad location data.
As someone who has contributed several articles to this "Confessions" series, beware that they're designed to highlight the worst-of-the-worst situation.
It's true that there are many junk vendors. Heavy politics and misaligned incentives for ad agencies usually means that the bad companies do better than the trusted vendors because fake data can obviously be shaped to look better than the real results.
That being said, location data does work. There are many ways to collect it from visual scanners in doorways, to open wifi networks that ping phones, to ISPs enriching data feeds. There are also 1000s of analytics SDKs embedded in apps that send pings constantly, so having a specific app open is not a necessity and never actually used by any serious network. Pretending that's the only way it works is just misleading.
One of the main promises of hyper-local geotargeting was the ability to build audience pools based off of location data that could be correlated with inferred home location mashed up with Zillow data to provide home value and estimated equity. On paper this kind of targeting would be really appealing to home improvement chains, banks, and even restaurants. I never really bought that way when I was a media buyer, so I can't say if it was ever effective. That said, I imagine it was more compelling to marketers when the data was sourced from a few trusted entities. Now, practically every free-to-play android app is asking for location data 24/7, no doubt to be sold to data exchanges. Once a location data relationship is established with an app publisher, there really is nothing stopping them from feeding psuedo-random location pings and getting paid for it. Its just another form of ad fraud. I wouldn't recommend a major digital buy to my worst enemy these days; Facebook lies about their metrics, Google sticks your ads next to child porn, display has been destroyed by domain spoofing and arbitrage. Maybe good old-fashioned direct buys with publishers are still safe, but there is a reason why TV budgets have held strong.
This just kicks the can one level down the road. Aggregator companies like PlaceIQ, Foursquare and Yelp will act as the intermediary and create more rigorous ways to detect spammy / useless location signals, and smooth out location tracking data.
Those intermediaries are the main worries, because brands or marketers may not trust every random Android app, but they will trust Yelp or Foursquare or something, and frankly given additional scale from data bartering and SDK pings that send back location from many other apps, places like Foursquare or Yelp really can provide targeting services that come close to what adtech has always promised.
To be clear, I’d consider this a bad thing, because places like Foursquare or Yelp, whatever they may say to put a PR spin on it, are quite literally preying on people who unwittingly share their location data and don’t really understand the terms, especially not when it’s some tertiary SDK traffic logs agreement causing some music app or hotel app or weather app to send them data tied to your IP address or device ID.
If those businesses can’t monetize their basic value proposition, like a app to search reviews of restaurants, it’s a signal to delete & shutdown the app... but unfortunately it’s become the reverse: a signal to abandon investment into the actual user and diversify all kinds of deceitful behind the scenes ways of getting user data and making users into the product.
I think we all know what the quote is about even though it's not an actual Ponzi scheme, it's just fraud. Another good reminder to question everything, especially group wisdom.
Adtech industry is about tools for things of questionable ethics. You shouldn't expect people doing and selling that to be paragons of honesty, who won't use their tools and methods against competition and customers within the industry.
Reminds me of this story about why whales don't die of cancer: because they're so large, that by the time their cancer grows to be dangerous, it gets its own cancer and dies.
The BBC article you posted looks like it has results that are 1) derived from actual animals, and 2) more directly useful, but I'm just a little disappointed that tumor cannibalism isn't their explanation, if I'm honest.
I don't think this hypothesis is widely accepted, but I remembered it because it's neat, and fits the game-theoretic model of cooperation/defection perfectly - which also makes it extremely useful for drawing analogies.
Tumors can be seen as made of cells which eschew cooperation with the rest of the body, and instead selfishly multiply, to the detriment of the whole and ultimately themselves.
It is my belief that advertising is a cancer on the society; it's exploiting - and in the process, destroying - every vulnerable individual and social heuristic. It involves uncooperative behaviors like manipulating people and lying to them.
The analogy here is that since entities making up the advertising industry eschew cooperation and embrace exploiting others for short-term gains, they're not going to magically start playing fair and cooperating within the industry. Therefore, to the extent you expect advertisers (including adtech) to scam you, they'll scam each other just the same - as seen in this article.
This, fortunately, somewhat limits the effectiveness of that industry.
I came up with this analogy few years ago, when I read accusations that Optimizely designed their A/B testing suite's UI in a way that promotes drawing statistically unsound conclusions from A/B tests, misleading you to believe that the tested intervention worked - and thus making you think Optimizely is successfully helping you learn things. My own personal observations from working alongside one social marketing team also confirmed the soundness of this analogy.
Location aggregation companies, like Yelp and Foursquare, actually do have large enough scale in terms of always-on location trail data to legitimately attempt this type of targeting. Combined with the fact that you’re giving them free data about your visits to places, your tastes, your likely income level based on where you shop, your home location, and from this they could reasonably predict your gender and put you into all sorts of specific advertiser buckets, it’s frankly perfectly reasonable to claim you could use it for ad targeting or measuring ad campaigns, even with a low match rate against the advertiser’s actual list of customers (under 1% surely).
To boot, Foursquare at least, and probably Yelp too, does data swapping and SDK agreements, like some thing recently announced with Accuweather (gross) and with Hilton Hotels.
I think people sincerely fail to imagine the real scope of this type of egregious trust violation and surveillance business model.
You can dress it up with whatever language you want about providing value to the user that makes them agreeable to the data collection terms, but I’m sure half of Foursquare users or Yelp users don’t actually know if they have the background location tracking disabled or not, or what other innocuous-seeming apps are silently feeding location pings to build a Foursquare data history about you to make you targetable for ads.
Frankly, I’d personally advise anyone to absolutely delete these apps or anything like them, and essentially vote with your wallet / vote by boycott and just refuse any apps that rely on this kind of business model.
so much true. people are worried about their data being sold to advertisers, while the little dirty secret of the ad-tech industry is that nobody has any data, tbh.
facebook and google do, but they are not selling it. and the rest of ad-tech loves to boast about their "data" on every conference while all they have is /dev/random output, more or less.
Well, yes, let's say that your flashlight app requested your location data and you agreed. You were in a park during dawn and you were looking for the stairs.
And there were like 2 ads shown to you during this 50 seconds period, while you were in the dark and not looking at your phone but using it as a duh, flashlight. And then you stopped the app to save your battery.
Do you really think some advertiser benefited by using this data?
And what "independent" ad-tech is dealing with is millions of cases above. Consistency is important. FB and GOOG have consistent data, collected over days and years, properly matched, correctly aggregated. All others are dealing with flashlight crap.
That's not how it works. Do you actually have adtech experience?
Mobile apps are 1000x more invasive than anything on the web and these SDKs constantly send data in the background. There are also 100s of signals that get correlated to build single device profiles across time. FB/GOOG do have an advantage but that doesn't mean everyone else is useless and if you have ISP partnerships then you can actually get even more consistent data down to a device and household.
A lot of those arguing your parent comment's point in other threads here and elsewhere seem to magically forget background data. Out of sight, out of mind.
Use a simple app to track background connections on your phone (e.g. NetGuard or the like) and you'd be amazed at how much gets sent back to the mothership.
An even dirtier secret is that you can have great data like Facebook but still serve incredibly irrelevant ads that everyone ignores. Google and FB presumably know massive amounts about me but I've never been shown an ad on either service that's worth clicking (on the off chance I even bother to look at them).
They can only sell the ads they've got in inventory. I've been on FB since 2006 and the last ads I remember seeing were for godawful Farmville-style games and shady dating sites. My profile would tell them I'm college-educated and living in a developed country with a network of similarly educated friends. And yet, I have never seen a single ad that didn't look like it would have fit right in at a late night infomercial.
The Twitter PWA Chrome app sadly doesn't block Promoted Tweets, and it's the same thing there; low-quality ads from garbage companies I'd never consider buying from.
People always say nothing works because they never see relevant advertisements.
But the disconnect here is the at the ad VIEWER is not the customer, the purchaser of the ad is the customer. It might be whoever has a relevant ad for you won't bid enough for you to see it.
That's a really interesting point.
A lot of people seem to see FB & Goog as adtech or something, when in fact they simply sell one of the main requirements for advertising - ad space.
Whether the ad is well targeted or not, the host still makes a profit (both monetarily and in terms of their own tracking data)
That's solid economics. If a war breaks out, be the one that sells guns. I seem to recall a quote along these lines but it escapes me...
Absolutely. They suck, they just suck less. With FB, you can _at least_ try to advertise to "small business owners in CA" and get a few leads for your SaaS.
Good luck doing it with independent ad-tech and their "data".
It really only plays as a branding message- like a superbowl ad. I've never bought anything immediately after seeing a superbowl ad, but the brand (Doritos, whatever) is in my brain now (as I eat my Tostitos of course).
I have a great deal of experience, and I've seen that you can generate final sale for some non-brand obscure SaaS platform (so no branding effect here at all) for like $30 per sale in Facebook, $90 in a good independent ad-tech platform, and $600 in a bad platform.
Facebook still totally kicks ass and absolutely can make money for you if your SaaS brings you $100 in a first year. (100-30=70 in incremental profit for you). Hell, even really good ad-tech platform (yes, they exist) can make you money, though less so (100-90=10 in a first year). Shitty platforms are money sink, though
True, but being old and slow telco businesses they totally failed to monetize it.
Verizon bought AOl, HuffPost, Techcrunch and plenty of other media to build an advertising giant by pairing data with sites audience -- and took 4.6 BILLION WRITEDOWN on the whole venture [1] because failed to make it work.
> Carriers sell information about you to data aggregators, which normally require the user to consent before selling it on further. But some third parties opted to sell information, like people's whereabouts, on to bodies such as bail bond companies, bounty hunters and landlords. In its original report, Motherboard paid an bounty hunter $300 to get the location of a phone to within a few hundred meters.
They totally failed to monetize it for advertising purposes. Taking cases like $300 per single person location, paid by bounty hunter, is arguing in a bad faith, and has nothing to do with ad-tech.
It will be interesting to see how the mobile operators now try to monitize their 1985 - 2007 location data collected b4 GOOG/FB were in the game, will they just sell it to Google?
In fact, this is almost tautologically true. If you have:
-- good data
-- on large enough scale
you are not going to sell it. You going to take advertiser dollars and match their ads against right audience, while safeguarding this data as your main know-how. Go and buy "barbell training fans in FL, US" and "women interested in fashion in CA, US" from facebook. Sorry, not for sale.
If instead you decide to "sell the data" it means that you can't really make it work for advertisers. Or, in other words, said data is mostly useless.
I don't know why someone would posit this falsehood.
You are going to sell it to EVERYONE you can (legally and sometimes not). There is no ad-tech company that does not take this stance. You get information about Honda buyers, you're going to sell it to Ford, then take what you sold to Ford (exclusively) and sell a repackaged format of the data (summary) you sold to Ford and sell that to Subaru ALONG with the data Ford asked for in the future.
Data management platforms (DMPs) are an entire concept built around selling the data. Every adtech company (that has any tech) sells the data that is collected, outside of using it for internal targeting. The number of demand partners and amount of useful targeting data to leverage for them, is necessarily a smaller set than what you collect.
Edit: This is experience from multiple senior ad-tech positions.
>I don't know why someone would posit this falsehood.
lol
>You are going to sell it to EVERYONE.
Exactly. And you know what you end up with? I checked lingerie website yesterday to buy a present for my girlfriend, this data was sold, now I'm a women in some DMP. I've bought a bottle of Talisker, now i'm wealthy man in the same DMP (even if I can only afford Talisker one time per year!!!). I've also bought some drugs for my grandfather, and now i'm +70 audience and I'm getting "senior discounts" ads while not being even 35 years old.
This is the problem with data platforms. Take data from everyone, mix it, and you have /dev/random. It's totally fine to milk advertisers for their money (which original article is all about), and not fine for anything else.
Advertisers finally start to understand it. And this is why GOOG and FB are getting all the money and ad-tech is slowly dying.
EDIT: And since -- as you correctly noted -- everyone is incentivised to sell ALL data to EVERYONE -- this is how you end up with a pool full of piss instead of water. Because everyone sold everything to this pool.
I think this is true, but there is rational for why you would sell the data even if the data is good. On a hypothetical scenario, say there was a great app for checking price comparisons for some esoteric product, or even say televisions. People download the app, and use it in Best Buy, etc. to compare tv prices. The location data therefore might be high quality/relevancy in one situation, but have low frequency and use to be valuable as a single source for advertising/selling to a hedge fund. Therefore, there would be hypothetical value to a data vendor aggregating and normalizing all these small sources into one feed.
In practice, the level and quality could vary significantly, and the difficulty to normalize the feeds could be much harder in practice. If you get data in aggregate from Foursquare and Yelp, potentially it can tell you enough to gauge foot traffic in restaurants, but to your point, it probably can't be used to build an advertising platform.
There are plenty of legitimate data providers. They're just overshadowed by shady and fake data vendors because they're cheaper and provider "better looking" results that help agencies look better and earn their bonuses.
Most companies would never have to go the lengths insurers go. They'll hire PIs to literally follow claimants around IRL to make sure they really do have a workplace injury. I know this because I worked at an office tasked with reviewing and annotating the videos they recorded.
Well, that was true back in 70's and this true now. If some brick-and-mortar business knew a lot about you (like your bank) -- it still knows a lot about you.
Ad-tech is different by boasting that they do know a lot about you, while in fact they have no clue.
The Google Opinion Rewards app seems to ask questions related to this all the time. Anybody know more about that?
5 out of 10 "surveys" go like this: Which of the following places did you visit recently? Yes? How did you pay? (Credit card, debit card, cash, made no payment, …).
To be honest 90% of the ad/location hypertargeting seems to be grasping at straws and much less effective than people think it is.
Retargeting/rebrokering it for ever smaller margins doesn't seem to make sense (except for those selling that snake oil).
Volume and campaign quality count more than just getting the exact words and the exact public (and remember you're paying more for a more targeted ad), targeted ads of course work for a targeted segment/audience, for most products that you would find in a high street/main street store, not really.
This is mostly pervasive because of agency/client relationships. Most clients go to an agency with a budget, and the agency, in good faith, is out there paying location data services to target their ads. Then a report is generated, passed to the client along some sort KPI that increased, and everyone is happy. This fraud exists because the market allows it.
Haha, total and utter bullshit. Someone is pulling a fast one on this sucker of a journalist. You don’t need location turned on to see where someone is lat/long. An average engineer can think of a dozen methods in half an hour.
You could've stated that better but yes, there are many ways to get location signal from mobile phones. Also people do have plenty of apps running and between all the SDKs embedded inside, there is location data being sent all the time.
Contrariwise, if location data from later subscribers were used to provide solid information to earlier subscribers that would be an effective and useful service.
If there is collusion to both sell fake data and detect fake data that is indeed fraud, but it's a different kind of fraud.