And Then You're Dead

The Permanent Record / Dossier · Data Brokers

The Segment You're In

A credit bureau at least admits it has a file on you, and the law gives you a way to ask for it. The industry described here has no such obligation for most of what it sells, no consumer-facing brand, and in the Federal Trade Commission's own words, no centralized place a person could go to find out which companies hold their data.

In December 2012 the FTC issued compulsory orders to nine data brokers under Section 6(b) of the FTC Act. The report it published in May 2014 remains the most detailed public accounting of the industry's mechanics. None of the nine collected data directly from consumers. Seven of the nine bought from or sold to each other. One held information on 1.4 billion consumer transactions and over 700 billion aggregated data elements, and another maintained over 3,000 data segments for nearly every U.S. consumer.

And Then You're Dead · October 2026 · 10 min read

Join the discussion →
A smartphone held in one hand displaying a navigation map application with a route drawn on it
A phone running a navigation app. The map is the service the user asked for. The stream of coordinates it produces is a separate product, sold to parties the user will never be told the names of. Photo: Santeri Viinamäki, CC BY-SA 4.0, via Wikimedia Commons.

The word "file" is misleading here. What a data broker holds about a person is not a dossier assembled by someone who was interested in them. It is a position in several thousand overlapping lists, most of which were generated by a model rather than observed, and most of which exist because somebody paid to be able to buy people who look like you.

The FTC's Section 6(b) authority lets it compel companies to file special reports without any allegation of wrongdoing. In December 2012 it used that authority on nine firms: Acxiom, CoreLogic, Datalogix, eBureau, ID Analytics, Intelius, PeekYou, Rapleaf and Recorded Future. The resulting May 2014 report, Data Brokers: A Call for Transparency and Accountability, is not an enforcement document. It is a description, produced under oath, of how the machinery works.

Nobody Asked You

The first finding is the structural one. None of the nine data brokers collect data directly from consumers. They collect from government sources — all but three of the nine obtained information directly from federal government sources — from commercial sources, from publicly available sources, and, crucially, from each other. Seven of the nine bought from or sold information to at least one of the other eight. The Commission observed that as a result, it would be virtually impossible for a consumer to determine how a data broker obtained a given piece of information about them, because the answer is usually another data broker.

The scale in the report is the scale of 2014, and it was already difficult to hold in the head. Acxiom's databases covered 700 million consumers worldwide with over 3,000 data segments for nearly every American. One of the nine held information on 1.4 billion consumer transactions and over 700 billion aggregated data elements. eBureau was adding over three billion new records a month. Intelius held more than twenty billion records. Rapleaf had at least one data point associated with over 80 percent of U.S. consumer email addresses.

Derived Data

What makes this a decision system rather than a mailing list is the second layer. Brokers store the actual data — name, address, home ownership, age, a magazine subscription, a purchase — and then derive more from it. The FTC's illustrations are deliberately mundane: a broker might infer that a person with a boating license is interested in boating, that a subscriber to Wired has a technology interest, that someone who has bought two Fords is loyal to the brand.

Those inferences get packaged into segments, built two ways. The simple way is combination: everyone with these attributes goes on this list. The other way is modelling — take a group known to have done something, find the characteristics they share, and apply those characteristics to everyone else. That is how a broker produces a segment called "Consumers Interested in Buying Camping Gear," and it is also how one produces a segment predicting which consumers are likely to seek a credit card chargeback.

The FTC's warning about these categories is the sharpest passage in the report, and it is about context rather than content. A segment called "Biker Enthusiasts" lets a motorcycle dealership send a coupon; an insurance company reading the same segment may infer risky behavior. A "Diabetes Interest" category lets a manufacturer of sugar-free products offer a discount; an insurer can read it as a health classification. The segment does not change when the buyer changes. Only the consequence does.

Some segments are not neutral even at rest. The report lists real ones: "Urban Scramble" and "Mobile Mixers," both described as containing a high concentration of Latino and African-American consumers with low incomes. "Rural Everlasting," single men and women over 66 with low educational attainment and low net worth. "Thrifty Elders." "Financially Challenged." "Underbanked Indicator." A person does not apply to be in these. They are placed there by a model, and the placement is a commodity.

0
data segments one of the nine brokers maintained for nearly every U.S. consumer, built from a mixture of observed and inferred attributes
0
of the nine brokers that bought from or sold data to one another, making the origin of any single fact effectively untraceable by the person it describes

Three Products, Three Different Sets of Rights

The nine firms sold into three markets, and what a person can do about their own data depends entirely on which one they have run into.

Five of the nine sold marketing products. Most gave consumers limited access to some but not all of the actual and derived data held about them. Only two allowed a consumer to correct their information for marketing purposes. Four of the five allowed an opt-out.

Four sold risk mitigation products — identity verification and fraud detection, the systems a lender consults to decide whether the person opening the account is who they say they are, or whether a Social Security number has recently been associated with too many addresses. This is the category with real consequences, and it is the category with the least recourse. Two of the nine gave consumers any access to the information used in these products, after verifying identity. One allowed correction. A person can be blocked from completing a transaction by a product in this category and, in the FTC's assessment, may not know it happened, may not know which company caused it, and may not be able to see or fix the underlying data.

Three sold people search products, the consumer-facing kind, where access and correction were comparatively good because the product is a search engine pointed at the person themselves.

Two further findings complete the picture. Some of the brokers stored all data indefinitely, even after it was updated, unless a contract said otherwise. And the choices that did exist were, in the Commission's phrase, largely invisible and incomplete: because data brokers are not consumer-facing, consumers may not know where to go to exercise them, and no centralized portal existed for consumers to learn about data brokers and what rights they offer.

The same segment means one thing to a motorcycle dealer and another to an insurer. The record does not change when the buyer does. Only what it costs you changes.

The Location Layer

Everything above describes the industry before the phone became the main sensor. Between January 2024 and December 2024 the FTC brought a run of cases that documented the newer version with the same specificity.

In January 2024 it settled with X-Mode Social and its successor Outlogic. X-Mode obtained precise location data three ways: through a software development kit that third-party app developers embedded in their own apps, through apps X-Mode itself published, and by buying from other brokers and aggregators. It sold to hundreds of clients across industries from real estate to finance, as well as to private government contractors. One documented use: supplying data on people who visited certain medical facilities in Columbus, Ohio and then went to pharmacies or specialty infusion centers, to a clinical research company for marketing. The order required the company to delete or destroy the location data it had collected and to build a program preventing recipients from associating its data with sensitive venues — medical and reproductive health clinics, places of worship, domestic abuse shelters, LGBTQ+ organizations, and political demonstrations.

Weeks later the FTC settled with InMarket Media, imposing what it described as its first ban on selling or licensing precise location data. InMarket had collected location through its own apps, including a shopping rewards app and a shopping list app, cross-referenced the histories against other personal data, and sorted people into nearly 2,000 audience segment lists with names like "parents of preschoolers," "Christian church goers," and "wealthy and not healthy." It retained the geolocation data for five years.

In December 2024 came the two that matter most for understanding the plumbing. Mobilewalla, the FTC alleged, obtained location data from real-time bidding exchanges — the millisecond auctions that decide which ad you see — and kept it even when its bid lost, which is to say it was harvesting the auction rather than participating in it. Between January 2018 and June 2020 it collected more than 500 million unique consumer advertising identifiers paired with precise location. It built segments from that, including one identifying people who attended the 2020 protests following George Floyd's death, analyzed by race and residency. The proposed order bars it from collecting data from advertising auctions for any purpose other than participating in them.

The same day, the FTC acted against Gravy Analytics and its subsidiary Venntel, which claimed to collect, process and curate more than 17 billion signals from around a billion mobile devices daily, obtained from other data suppliers rather than from consumers, for what the Commission described as commercial and government uses. The proposed order requires deletion of the historical location data, a sensitive location program covering eight categories including medical facilities, religious organizations, correctional facilities, labor union offices, schools and childcare, services supporting people by racial or ethnic background, shelters for homeless, domestic abuse, refugee or immigrant populations, and military installations, and a supplier assessment program to verify that consumers consented in the first place.

What Recourse Looks Like Now

The 2014 report's central complaint — that no centralized portal exists for a consumer to learn who holds their data — has been partially answered, in one state. California requires data brokers to register annually with its privacy regulator between January 1 and 31, reporting their data collection categories and who they share with. Under the same law the state built DROP, the Delete Request and Opt-out Platform, which went live for consumer requests on January 1, 2026 and which registered brokers must begin processing from August 1, 2026. One request, submitted once, is meant to reach every active registered broker.

That is a real mechanism and it is worth using. It is also a state registry, covering companies that register, reaching the ones that comply, for residents of one state. The nine firms the FTC studied were the visible end of an industry whose size, the Bureau of Consumer Financial Protection would later note in a different market, is unknown because nobody counts it.

And Then You're Dead

The credit file at least has a statute attached to it, a dispute address, and a legal fiction that it is about you specifically. The broker record has none of that. It is assembled from purchases you made, records the government published, signals your phone emitted while you were asleep, and inferences drawn about people who resemble you; it is bought and resold among firms whose only relationship to you is that they have never met you; it is retained indefinitely; and its most consequential use — the risk product that quietly declines your transaction — is the one where almost nobody can see the data, and almost nobody can correct it. You are not in a file. You are in about three thousand lists, and you did not put yourself on any of them.

That's it. That's the whole thing.

Comments

Loading…

Comments are anonymous and moderated.

Sources