ROR Labs cover: 4 points identify you. Four approximate time-and-place readings single out 95% of people in a mobility dataset. That is why your location is worth selling.
|

Four Points Identify You. That Is Why Your Location Is Worth Selling.

Further reading: Means of Control — Byron Tau (Crown, 2024). A Wall Street Journal reporter’s investigation into location-data brokers and government surveillance purchasing — narrower than Zuboff’s argument, but more specific about how the data actually moves. Find it on Amazon (paid link)

Disclosure: this article contains affiliate links, marked (paid link). If you buy through one we may earn a commission, at no extra cost to you. As an Amazon Associate I earn from qualifying purchases. It costs you nothing and it does not change what we recommend.

Key takeaways · 14 min read

  • Four known places-and-times identify 95% of people in an anonymised mobility dataset. “Anonymised” means the name was not attached, not that it cannot be.
  • Coarsening the data barely helps — uniqueness decays as roughly the 1/10 power of resolution.
  • The main pipe is real-time bidding. A bid request broadcasts your device ID and location to many bidders at once.
  • Losing the auction does not lose the data. The FTC found a broker retaining what it received from auctions it never won.

On 1 August 2026, California switched on something that has never existed anywhere before: a single web form where a resident can tell every registered data broker at once to delete what they hold. It is called DROP, brokers have to check it at least every 45 days, and it went live three weeks ago to almost no coverage.

It exists because of a specific problem that the industry spent a decade insisting was not a problem. The data being bought and sold is described as anonymous: no name attached, just an advertising identifier and a string of coordinates. The claim is that this makes it harmless. The research says otherwise, and it has said otherwise since 2013.

This article is about where that data actually comes from — the answer is stranger than most people expect, and it involves an auction you are entered into hundreds of times a day — what regulators found when they finally looked inside two of these companies, and which of the available defences do anything. It is also, unusually for this site, an article with almost nothing to sell you.

A stone bridge of arches crossing a river at dusk, its lamps lit, the city low on the far bank and everything doubled in the water.
Everyone crosses here. That is exactly what makes it useful.

Four points are enough

The foundational study here was published in Scientific Reports in 2013 by de Montjoye, Hidalgo, Verleysen and Blondel. They took fifteen months of mobile phone mobility data for about 1.5 million people, stripped of names, with location recorded at the level of the cell antenna and time recorded to the hour — roughly the crudest useful resolution there is.

Then they asked how many known points about a person it takes to pick their trace out of the crowd.

Four spatio-temporal points identify 95% of people

Anonymised mobility traces, 1.5 million individuals, 15 months, hourly resolution at antenna level.

4known places-and-times — where you were, roughly when — are enough to uniquely identify 95% of individuals in the dataset
1/10the power by which uniqueness decays as resolution is coarsened. Blurring the data barely helps.
Four points is not a lot. Home and work are two of them and are trivially inferable from the trace itself — where the device sits overnight and where it sits on weekday afternoons. A photo you posted, a shop loyalty record, a friend mentioning where you had lunch: that is the other two. The decay result is the part that matters most for policy, because it means the standard industry fix — round off the coordinates, aggregate the timestamps — does not restore anonymity. It just costs you accuracy.

Source: de Montjoye, Y.-A., Hidalgo, C.A., Verleysen, M., Blondel, V.D. “Unique in the Crowd: The privacy bounds of human mobility.” Scientific Reports 3, 1376 (2013).

So when a company says its location dataset is anonymised, the accurate translation is: we did not attach the name ourselves. Whether the name can be reattached is a different question, and the answer is usually yes, by anyone with a modest amount of outside information. That is the entire foundation of the industry described below.

The auction you are entered into all day

A single street lamp at dusk casting a bright pool of light on a dark pavement, with a bin at the edge of it.
You were here. That is the whole of it.

Most people assume location data leaks because an app they installed sells it. That happens, but the larger and less intuitive pipe is real-time bidding — the mechanism behind almost every advert you see in an app or on a web page.

When an app has an advertising slot to fill, it broadcasts a bid request to an exchange. That request describes the opportunity so bidders can decide what it is worth: device identifier, app, and often precise location. The exchange forwards it to many potential buyers simultaneously. One of them wins and shows an advert. All of the others still received the data.

Losing the auction does not mean losing the data

The FTC found one broker harvesting from auctions it never won.

app on your phone ID + location ad exchange winning bidder shows you an advert losing bidder keeps the data anyway losing bidder keeps the data anyway
This is the part that surprises people. A bid request is not a transaction, it is a broadcast. A company can sit in the exchange, bid on nothing, win nothing, pay nothing, and collect the location of everyone whose bid requests pass through — which is how a firm you have never installed anything from ends up holding your movements.

Source: Federal Trade Commission complaint against Mobilewalla, December 2024.

What regulators found when they looked

Two U.S. Federal Trade Commission cases opened this industry up, and they are worth reading as descriptions of ordinary practice rather than as scandals. Neither company was accused of a breach. Both were doing what the business does.

Two enforcement actions, two years apart

Both settled with orders restricting the sale of sensitive location data — the first of their kind.

X-Mode Social / Outlogic (Jan 2024)Mobilewalla (Dec 2024)
How the data was gatheredIts own apps — including a nightlife app called Drunk Mode — plus a software kit embedded in other developers’ apps, plus purchases from other brokersReal-time bidding exchanges and third-party aggregators, retaining data from auctions it did not win
ScaleSold to hundreds of clients across real estate, finance and government contractingMore than 500 million unique advertising identifiers paired with precise location, January 2018 to June 2020
What the data revealedVisits to medical and reproductive health clinics, places of worship, domestic abuse shelters, LGBTQ+ locations and protestsVisits to pregnancy centres used to build a “pregnant women” segment; attendance at George Floyd protests, with inferred racial background and home city
What the order requiresStop selling sensitive location data; delete what was collected; verify upstream consent; provide deletion and opt-outNo sale or use of location tied to military sites, churches, healthcare and correctional facilities, union offices, political gatherings or LGBTQ+ locations; no collecting auction data for other purposes; provide deletion
Note what the second row is describing. Five hundred million advertising identifiers is not a targeted operation against particular people. It is the ambient exhaust of ordinary phone use, collected at national scale by a company most of its subjects had never heard of.

Sources: Federal Trade Commission press releases and complaints, “FTC Order Prohibits Data Broker X-Mode Social and Outlogic from Selling Sensitive Location Data” (January 2024) and “FTC Takes Action Against Mobilewalla for Collecting and Selling Sensitive Location Data” (December 2024).

The FTC chair’s framing of why this matters is the clearest short statement of the problem: geolocation data can reveal not just where a person lives and whom they spend time with, but which medical treatments they seek and where they worship.

Why the “sensitive place” frame is too narrow

The orders above work by listing categories of place: clinics, churches, shelters, union offices. That is a sensible legal instrument and an incomplete model of the risk, because most of what a location trace reveals is not a single sensitive visit. It is a pattern, and patterns are made of entirely ordinary places.

What an ordinary trace discloses without visiting anywhere sensitive

None of the inferences below require a protected location. All of them fall out of the pattern.

WHERE YOU LIVEThe place the device rests between roughly midnight and six. This is the single easiest inference in the dataset and it is what turns an identifier into an address.
WHERE YOU WORKThe weekday daytime cluster. Together with the first, this is two of the four points the 2013 study needed.
WHO YOU ARE WITHCo-location: two identifiers repeatedly in the same place at the same time. Relationships, affairs, business meetings and journalistic sources all show up as co-location patterns.
WHAT CHANGEDA trace that stops going to work, starts visiting one address weekly, or begins a new overnight location tells a story about illness, separation or job loss without naming any of them.
WHAT YOU EARNHome location alone is a strong proxy for income, and income proxies are what drive differential pricing and lending offers.
WHERE YOU WILL BEHuman movement is highly regular. A few months of history predicts next Tuesday well enough to be worth money.
This is why “I have nothing to hide” misses the shape of the problem. The question is not whether any single place you go is embarrassing. It is that a complete movement history is a different kind of object from the sum of its stops.

Sources: de Montjoye et al., Scientific Reports (2013); FTC complaints against Mobilewalla and X-Mode Social/Outlogic.

What actually helps, in order

The honest summary is that individual action here is weaker than in most subjects this site covers, because the collection is structural. That is not a reason to do nothing — the measures below genuinely reduce what is emitted — but anyone promising you disappearance is selling something.

Ranked by how much data each one stops

Every item on this list is free.

1. TURN OFF THE ADVERTISING IDBoth major phone platforms let you disable the advertising identifier entirely, in settings, in about fifteen seconds. It is the key that links auction records together over time. Without it the exhaust is far harder to assemble into a trace.
2. SET LOCATION TO “WHILE USING”Go through the location permission list app by app. Almost nothing needs background location. A weather app needs a city, not a continuous trail.
3. GIVE APPROXIMATE, NOT PRECISEBoth platforms now offer a coarse location option. For weather, news, retail and most social apps it works identically and removes the precision that makes a trace unique.
4. DELETE THE APPS YOU DO NOT USEThe most reliable way to stop an app emitting bid requests is for it not to be running. Games and utilities with heavy ad loads are the most prolific senders.
5. USE THE FREE LEGAL ROUTECalifornia residents can submit one deletion request to DROP that reaches every registered broker, from 1 August 2026. Several other states and the EU and UK give you a deletion right you can exercise directly.
6. AUDIT PERMISSIONS TWICE A YEARApps regain permissions at reinstall and after some updates. Both platforms now report which apps used location recently — that report is the useful one.
7. SEPARATE THE SENSITIVE TRIPIf there is a journey you do not want in a dataset, the only reliable answer is to not carry the device on it. Everything else is mitigation.
8. TREAT LOYALTY CARDS AS LOCATION DATARetail transaction records are one of the outside sources that reattach a name to an anonymous trace. They are collected under a different name and they solve the same problem for a buyer.

Sources: California Privacy Protection Agency, DROP; Federal Trade Commission consumer guidance; platform privacy settings documentation.

The California mechanism deserves a note of its own, because it is genuinely new and it is the first thing on this list that acts on data already collected rather than on data yet to be emitted.

DROP: one request, every registered broker

California’s Delete Request and Opt-Out Platform, operated by the state privacy agency.

1 Aug
2026
the date consumers could begin submitting deletion requests through a single verified form
45
days
the maximum interval at which every registered data broker must check the platform and process the requests waiting there
Registration is annual, in January, with a fee, and brokers must file disclosures about what they collect and sell. Transparency reports follow in 2027. It is the first regime anywhere that puts the burden of finding the consumer on the broker rather than the burden of finding hundreds of brokers on the consumer.

Source: California Privacy Protection Agency, Delete Request and Opt-Out Platform (DROP) and Data Broker Registry; California Delete Act (SB 362), as amended by SB 361.

The shortest shopping list on this site

There is very little here that a purchase fixes. One item earns its place, and not for the reason people buy it.

SIGNAL-BLOCKING POUCHA lined pouch blocks radio in and out. Against location brokers it is a narrow tool — useful only for a specific journey you deliberately leave the phone out of. The stronger everyday case is a different risk entirely: keeping a car key fob in one at home defeats relay attacks, in which thieves amplify the fob signal from inside your house to open and start the car on the drive.
Browse on Amazon →

Three things we are deliberately not linking, and this is the section that matters.

VPNs marketed as location privacy. A VPN hides your IP address from websites. It does nothing whatsoever about GPS, and an app you granted location permission to reads the satellite fix directly — the tunnel is irrelevant. This is one of the most heavily advertised false claims in consumer technology and the affiliate commissions on it are large, which is presumably related.

Paid data-removal subscriptions. They send opt-out requests to brokers on your behalf, which is a real service badly matched to this particular problem: they reach the brokers that publish an opt-out and comply with it, and those are the ones you could have reached yourself. California residents now have DROP, which reaches every registered broker for nothing. Elsewhere the honest position is that these services buy convenience, not coverage, and they should be sold that way.

“Anti-tracking” phone cases and privacy gadgets. A camera cover is fine. A case that claims to stop tracking is either a signal blocker, in which case your phone is off and you knew that, or it is nothing.

A car park seen from above: painted bays running at an angle, five cars parked, one bay empty with a trolley left in it.
Four of these, on four days, and it is no longer anonymous.

Questions people ask

Does turning off location services solve this?

It removes the precise fix, which is the most valuable part, and it does not make you invisible. Your network operator knows which cell tower you are attached to as a condition of the phone working at all, and IP addresses carry coarse location. What you are removing is the metre-accurate, timestamped trail that makes a dataset uniquely identifying — which is most of the value, so it is worth doing.

Is this legal?

Largely, in the United States, which is why the FTC actions were brought under unfairness and deception rules rather than a privacy statute — the theory was that consumers had not meaningfully consented, not that collection is banned. In the EU and UK, the GDPR treats precise location as personal data and much of this practice sits on far weaker legal ground, which is one reason the datasets and the enforcement look different there.

I am not in California. Is DROP useless to me?

Directly, yes — it is a California resident right. Indirectly it may not be, because compliance systems built for one large market tend to get applied more broadly, and because it is the first working template for the rest. Meanwhile a growing number of U.S. states, and the EU and UK, give you an enforceable deletion right you can exercise against a named broker yourself.

Should I worry about this more than about scams?

No, and we would rather say so. Measured in money lost this year, fraud is a far larger and more immediate risk than data brokerage — that is the subject of our pieces on credit freezes and voice-cloning scams. Location brokerage is a slower risk with a different shape: it rarely costs you a specific sum on a specific day, and it is much harder to undo once the data exists.

What about my children’s phones?

The same settings apply and matter more, because free games carry heavy advertising loads and a child’s trace covers a school, a home and a routine. Disable the advertising identifier on the device, set location permissions to “while using” and approximate, and treat any app that demands background precise location for no reason as a reason to choose a different app.

The short version

  • Four known places-and-times identify 95% of people in an anonymised mobility dataset. “Anonymised” means the name was not attached, not that it cannot be.
  • Coarsening the data barely helps — uniqueness decays as roughly the 1/10 power of resolution.
  • The main pipe is real-time bidding. A bid request broadcasts your device ID and location to many bidders at once.
  • Losing the auction does not lose the data. The FTC found a broker retaining what it received from auctions it never won.
  • Mobilewalla held 500 million+ advertising identifiers paired with precise location, and built segments from visits to pregnancy centres and protests.
  • X-Mode gathered through its own apps and a kit inside other developers’ apps, and sold to hundreds of clients including government contractors.
  • The real disclosure is the pattern: home, work, who you are with, what changed — none of which requires a sensitive destination.
  • Free and effective: disable the advertising ID, set location to while using and approximate, delete unused apps.
  • California’s DROP opened on 1 August 2026; brokers must check it every 45 days.
  • A VPN does nothing about GPS. Paid removal services buy convenience, not coverage.

This article describes data practices and consumer rights as of August 2026. It is not legal advice, privacy law differs sharply by country and state, and the rules here are changing faster than almost any other subject on this site — check the current position with your own regulator before relying on any of it.

On the links above: some are affiliate links, marked (paid link). If you buy through one we may earn a commission at no additional cost to you. As an Amazon Associate I earn from qualifying purchases. We link to product searches rather than specific items so that recommendations do not break as models change, and we say plainly when we are choosing not to link something. Full policy: Affiliate Disclosure.

Sources

  • de Montjoye, Y.-A., Hidalgo, C.A., Verleysen, M., Blondel, V.D. “Unique in the Crowd: The privacy bounds of human mobility.” Scientific Reports 3, 1376 (2013). (1.5 million individuals over 15 months; four spatio-temporal points identify 95%; uniqueness decays as the 1/10 power of resolution.)
  • Federal Trade Commission. “FTC Takes Action Against Mobilewalla for Collecting and Selling Sensitive Location Data,” December 2024, and the finalised order, January 2025. (500 million+ advertising identifiers with precise location, January 2018 to June 2020; retention of real-time bidding data from unsuccessful bids; pregnancy centre and protest segments.)
  • Federal Trade Commission. “FTC Order Prohibits Data Broker X-Mode Social and Outlogic from Selling Sensitive Location Data,” January 2024, and the finalised order, April 2024.
  • California Privacy Protection Agency. Delete Request and Opt-Out Platform (DROP) and Data Broker Registry. (Consumer requests from 1 August 2026; brokers must access the mechanism at least every 45 days; annual registration each January.)
  • California Delete Act (SB 362, 2023), as amended by SB 361 (2025).
  • Identity Theft Resource Center. 2025 Annual Data Breach Report.

Similar Posts