As technology advances, each new development often builds on what came before. The growing flood of data from connected devices has opened new opportunities for companies to target people with increasing precision. Over the past century, advertising and pricing have followed this trajectory, moving through four distinct phases: from broad, mass-market strategies to today’s highly personalized systems rooted in commercial surveillance.

Phase 1: Mass media and crude group segmentation (1920s – 1980s)

This first period features a period of railroad expansion and widespread automobile adoption. At the same time, businesses found new methods of reaching people directly, bypassing traditional intermediaries like local stores. This includes direct mail catalogs, radio broadcasts and television programs and ads – all powerful new tools to reach millions of consumers directly in their homes.

Audience segmentation was crude but growing. Devices like the Audimeter, a black box meter attached to radios and TV sets,  sampled listening and viewing habits, creating quantifiable segments for marketers. Surveys from companies like Nielsen enabled targeting by demographic variables such as age, income, gender, and zip code. Local radio and TV affiliates sold advertising spots, and companies experimented with sponsoring content rather than interrupting it. Early A/B testing occurred through couponing, promotions, and broadcast offers.

Pricing strategies also reflected segmentation. Railroads employed monopolistic practices and used crude group segmentation to target small town farmers with higher rates and fewer discounts than larger corporations on routes, storage fees, and products. Congressional reports documented these practices, ultimately leading to legislation like the Interstate Commerce Act, a reminder of how crude segmentation could influence prices long before digital tracking.

Department stores like Montgomery Ward and Sears used mail catalogs into direct lines to the American home. Data brokers emerged, collecting names, addresses, and personal details and selling them to retailers. As a 1944 Saturday Evening Post article noted, a business could rent a list to reach “left-handed golfers, dentists, ministers,” revealing how even early targeting relied on personal data, however rudimentary.

Phase 2: More granular, identity based targeting (1980s – 2000s)

The rise of the internet, search, personal computers, and early e-commerce marked the second phase. Companies like Google, Amazon, eBay, AOL, Yahoo!, and Hotmail transformed how people found information and shopped while customer relationship management (CRM) systems like Salesforce tracked purchases, complaints, and preferences. These new online behaviors generated a wide range of tracking techniques designed to enable greater analytics and precision on serving ads or pricing. Among common techniques include web tracking cookies,^[1] browser fingerprinting, web analytics tools, online customer panels, banner ads, contextual targeting, log-based analytics, and basic online A/B testing.

To see how this new wave of identity-based targeting played out in terms of prices and advertising, consider two early examples from Amazon.

In the early 2000s, online shoppers uncovered a confusing trend in an internet forum: the same DVD was being sold on Amazon at different prices to different customers. The differences, Amazon claimed, were done on a “totally random basis” – but in fact were testing how changes in price would affect sales volume and revenue, Amazon said. The company denied using demographics to change prices, but this instance is one of the first public glimpses into early e-commerce experimentation – changing prices for the same product to different people to maximize profit.

Roughly around a similar time, Amazon launched “item-based collaborative filtering,” a technique used to make recommendations to users based on their preferences. Using individual behavioral data, large-scale data processing, and real-time personalization – the company could generate real-time suggestions. “Customers who bought this item also bought…” prompts came with a row of recommended products based on their purchases and items they browsed in the store. Alternatively, in the final screen of a checkout online, more recommendations may appear. This technique was so effective that a Microsoft Research report estimated that “30 percent of Amazon.com’s page views were from recommendations.”

Phase 3: Social Media Platformization (2000s – 2020)

The first two decades of the twenty-first century witnessed a rapid proliferation of major tech companies that launched including Facebook, YouTube, Twitter, WhatsApp, Instagram, and Snapchat. Alongside the explosion of social media came recommendation systems, which transformed how people discover information on platforms like Netflix, Spotify, and LinkedIn’s “People You May Know,” tailoring information to people’s tastes and habits. At the same time, first generation cloud service providers – like Amazon Web Services, Google Cloud Platform, and Microsoft Azure – rewired how businesses accessed technology, shifting from on-prem servers to scalable digital infrastructures. In roughly a decade, these platforms became global enterprises and data powerhouses.

Social media companies were “free” to users in exchange for vast amounts of granular data. Every interaction – mouse movements, scrolling, dwell time, likes, comments, keystrokes, photo sharing, video watching, and social connections – feeds algorithms that adapt content and deliver highly targeted ads. Mobile devices amplified tracking, capturing precise geolocation data throughout daily activities. Beyond a stationary personal computer, with connected devices in their pockets, people can be tracked throughout their day – commutes, shopping trips, health visits, or even places of worship. Meanwhile, the Internet of Things added connected home devices—vacuum cleaners, thermostats, fitness bands, and security systems—further expanding data collection.

Third-party data brokers and third-party trackers – from early cookies to modern pixels, SDKs, or fingerprinting – became a central part of this commercial surveillance infrastructure. Operating largely unseen, these intermediaries collect, aggregate, and sell detailed profiles on individuals, often without their knowledge or consent. Real-time bidding reconfigured how ads were purchased and sold in milliseconds while cross-device tracking stitched together identities across mobile, desktop, and connected devices.

During this period, the rise in data-driven personalization was not just theoretical – it played out in concrete ways across pricing and advertising.

In 2014, Northeastern researchers analyzed the accounts and cookies of over 300 users to study price steering and discrimination on 16 popular e-commerce sites, including Orbitz, Expedia, Hotels.com, Home Depot, Travelocity, and Priceline. The findings revealed price discrimination, A/B testing to steer users toward more expensive hotels, and personalized search results based on mobile device use, clicks and purchases.

Similarly, a 2012 New York Times article revealed that Target could predict expectant mothers’ needs, sending targeted coupons for baby products based on subtle purchasing patterns, such as buying unscented lotion or certain supplements like calcium, magnesium and zinc. These developments illustrated how mass data collection accelerates companies’ ability to target ads and set individualized prices.

Phase 4: Commercial Surveillance (2020 – Present)

The early 2020s ushered in a new chapter of commercial surveillance, defined by the rise of generative AI. Although the roots of generative AI can be traced back to mid-20th century research, the defining moment came in late 2022 when OpenAI launched ChatGPT – reaching one hundred million users in a staggering two months. What followed was the widespread deployment of large language models embedded into everyday life, including search engines, virtual assistants, productivity tools, and systems that could generate text, audio, and video.

To refine these systems, companies began collecting mass amounts of data – both real and synthetic. AI agents are increasingly performing complex tasks on behalf of users like schedule management, customer support, and software development – though how effectively they do so remains an open question. Meanwhile, a dense collection of connected devices – cars, wearables, and televisions – stream data to companies, often through default settings that remain invisible to users. With sensors, cameras and wearable devices, digital and real-world spaces are increasingly blurred, capturing behavior and responses in real-time.

Beyond clicks and searches, companies have more ways to collect more precise and personal data – ranging from biometrics to behavioral cues like keystrokes or cursor movements. Firms can use machine learning models to make user inferences based on sensitive traits and behaviors, such as emotional states through eye gaze or tone, or assess risks related to health and creditworthiness. Firms claim they can predict how much a consumer might be willing to pay, how likely they are to buy, or whether they represent a financial risk. Importantly, rather than solely depending on third-party trackers and data brokers, companies can rely on first-party data collection directly through their own products and services – an arrangement that can provide rich, more continuous insight, but with little transparency or external oversight.

These dynamics become clearer when looking at how they unfold in practice.

Consider the case of rideshare platforms in terms of worker wages. To understand what portion of each fare actually made it into drivers’ pockets, Dana Calacci, Assistant Professor at Penn State, built a tool to crowdsource rideshare data from drivers themselves. Over an 18-month period, 45 drivers contributed data across 76,000 rides. The results were striking: rideshare platforms withheld 20-30 percent of each fare before paying drivers. This take-rate is opaque to drivers, obfuscating the fees and deductions – including service fees, convenience fees, bonuses, and profit margins – that are involved in the wage calculation.

The same lack of transparency surfaces in advertising. In 2023, regulators challenged how social media companies collect and exploit user data. The European Union fined Meta by $1.3 billion for transferring European users’ personal data to the U.S. without adequate protection which violated GDPR. The same year, Norway banned Meta from tracking users for targeted advertising without consent and South Korea fined Meta $15 million for illegally collecting sensitive information from Facebook users, including sexual orientation and political beliefs, without consent. Together, these cases illustrate how tech companies continue to profit from personal data.

Four insights on the evolution of targeting

These four phases trace how companies have learned to target consumers for over a century. What began as crude segmentation through mass media has transformed into an infrastructure capable of tailoring ads and prices to individuals in real time.

1. Advances in tracking have increased the resolution of consumer data. Beyond purchases or device type, companies log micro-level behaviors—such as mouse movements, scrolling speed, pauses on a page, or time spent hovering over content—capturing consumer actions with extensive granularity.
   
   
2. The scale of data collection has expanded dramatically. At the same time, the scale of data collection has expanded. Firms now draw from a wide array of behavioral signals – search queries, voice commands, biometric inputs, geolocation, and even subtle interaction patterns. This breadth of data allows them not only to tailor ads, but also to infer sensitive traits such as creditworthiness, health risks, or employability.
   
   
3. Mass collection is designed to operate by default, and is largely unavoidable. The data collection happens, for the most part, in a covert way – making a person’s ability to opt-out difficult. People rarely know why they are seeing what they see or paying what they pay. Or if there is an option to explore more, the details may be buried deep in legal disclosures.
   
   
4. Over time, these systems are adapting, capable of modifying ads and prices in near-real time. What’s more is that these systems can operate and update continuously – and can deploy more quickly, especially with the roll-out of more digital price tags in stores. Targeting for ads or prices can adapt within seconds, optimizing based on consumer actions, behaviors, and context.
   

Advertising and pricing tactics have quietly converged. The evolution of consumer targeting has been less a series of isolated innovations than a steady convergence, as data collection, personalization, and pricing strategies have fused into a single, always-on system of commercial surveillance. The tools and infrastructure built to target ads – including pixels, cookies, and software development kits – can also be used to power and determine the prices people pay. These phases reveal not just technological advancement, but a fundamental shift in how markets interact with individuals, turning data into power and transforming everyday purchases into opportunities for precise, targeted influence.

This historical mapping traces the evolution of advertising and pricing strategies, illustrating how each phase laid the groundwork for today’s price targeting. Part 2 of this series builds on this foundation, highlighting the critical areas where rigorous research and careful investigation are urgently needed to understand the implications and limits of these practices.

***

Stephanie T. Nguyen is a Senior Fellow researching the intersection of technology, artificial intelligence and regulation at the Vanderbilt Policy Accelerator and the Georgetown Institute for Technology Law & Policy. She was previously Chief Technologist and led the Office of Technology at the Federal Trade Commission. This post is adapted from a keynote speech delivered in July 2025, in Washington, D.C. at the Privacy Enhancing Technologies Symposium.

---

^[1] Cookies are text files with pieces of data that are used to identify your computer or a user as you explore the web.