Increasing Resolution: The Evolution of Advertising Measurement

For as long as businesses have tried to influence what people do, they have faced the same basic problem: How do you connect an action taken by the business with a behavior that happens later?

An advertisement appears.
A coupon is distributed.
A promotion launches.
An influencer mentions you.

Someone eventually visits, returns, buys, clicks, or does nothing at all.

The distance between these two moments has always been difficult to see clearly. Much of the history of advertising measurement is the history of trying to reduce that distance — increasing the resolution and speed with which businesses can observe what happened between action and response.

---

Significant Historical Milestones

➖ Before mass markets — proximity

Producer, seller, and buyer were often much closer together. Feedback could be direct and local.

Industrial production and expanding distribution increasingly separated the producer from the individual buyer. Mass advertising helped manufacturers reach those buyers. But that distance created a new problem: how could a manufacturer know which advertising had actually worked?

➖ Late 1800s — measurable response

Mass-produced branded goods and mass advertising expand. Coupons, mail offers, order forms, and similar devices begin leaving countable responses. Coca-Cola’s large-scale coupon campaigns illustrate this early shift toward promotions that could produce a visible, countable response.

➖ Early 1900s — source identification

Keyed coupons/order forms allow advertisers to ask not merely did somebody respond? But also, which advertisement did they respond to?

⏵ Broadcast era — audience without individual traceability

Radio and television enormously expand reach, while direct connection between exposure and purchase becomes harder. Audience measurement tells advertisers who probably encountered something; sales tell them what happened afterward. Connecting the two remains difficult.

➖ 1950s–1980s — modeling effects

Statistical/econometric approaches develop to estimate relationships among advertising, promotions, sales, seasonality, and other variables. Marketing-mix modeling eventually becomes an important approach for channels where individual-level tracking isn't possible.

⏵ Direct marketing, barcodes, loyalty systems — individual traces grow richer

Responses, purchases, households, products, promotions, and eventually customer identities become increasingly linkable. Coupon clearinghouses and UPC scanning are useful milestones here; Nielsen established a coupon clearinghouse in 1957, and UPC scanning appeared in retail in 1974.

➖ 1990s–2000s — the click changes expectations

Digital advertising can increasingly connect impression → click → visit → action, producing the feeling that attribution can become deterministic rather than inferred.

➖ 2010s — multi-touch everything

Multiple devices, channels, platforms, retargeting, identity graphs, and customer data make the trace richer while simultaneously making the causal question harder: which touch actually mattered?

➖ 2020s — the attribution paradox

We have vastly more data and vastly more sophisticated models, but privacy restrictions, disappearing identifiers, fragmented platforms, and increasingly complex journeys have renewed interest in aggregate experimentation and MMM. More traces have not eliminated the fundamental problem of causation.

---

What Advertisers Can Connect Today

Today’s advertising systems can draw from a much richer collection of signals than the early coupon ever provided. Not every advertiser has access to all of them, and privacy rules, platforms, permissions, and technical architecture determine what can actually be used. But together they illustrate how dramatically measurement resolution has increased.

• Cookies: recognize a browser or session across interactions and, depending on the cookie, over time.
• Tracking pixels: create a record that an event such as a page view, email open, or conversion occurred.
• Loyalty IDs: connect transactions and other activity to a persistent customer relationship.
• Advertising IDs: provide mobile devices with pseudonymous identifiers that can support advertising measurement and targeting.
• Device and browser fingerprints: infer a distinctive signature from combinations of device or browser characteristics rather than relying on a single issued identifier.
• Identity graphs: connect multiple identifiers, devices, accounts, and signals into an inferred representation of the same person or household.
• Geolocation: add movement and place to other behavioral traces.
• Re-identification and de-identification: explore when supposedly anonymous or pseudonymous data can — or cannot — be associated again with a person.

---

Sources:

• Duke University & Library of Congress. (2001). The Emergence of Advertising in America, 1850–1920. Library of Congress. | Provides historical grounding for the emergence of mass advertising, consumer culture, branded goods, coupons, trade cards, and the professionalization of advertising.
• Hopkins, C. C. (1923). Scientific Advertising. | Documents early practices in measurable advertising, including keyed advertisements, traced returns, coupons, testing, cost per response, and comparison between advertising executions.
• Nielsen. (n.d.). Our History of Innovation. | Traces the development of systematic audience and promotional measurement through radio measurement, television audience indexes, local-market measurement, and coupon clearing.
• Campbell, R. H. (1965). “A Managerial Approach to Advertising Measurement.” Journal of Marketing, 29(4). | Shows advertisers confronting the measurement problem directly by the 1960s and developing more systematic approaches for evaluating advertising results.
• National Museum of American History, Smithsonian Institution. (1999). Behind the Lines: The Universal Product Code (U.P.C.) at 25. | Documents the retail-data transition represented by the first UPC scan in 1974 — a significant increase in the resolution with which individual products and transactions could be recorded.
• Lewis, R. A., Rao, J. M., & Reiley, D. H. (2015). “Measuring the Effects of Advertising: The Digital Frontier.” In Economic Analysis of the Digital Economy. University of Chicago Press. | Examines the shift toward digital measurement, including clicks, increasingly detailed behavioral data, experimentation, and the distinction between observable activity and causal advertising effects.
• Kannan, P. K., Reinartz, W., & Verhoef, P. C. (2016). “The Path to Purchase and Attribution Modeling: Introduction to Special Section.” International Journal of Research in Marketing, 33(3), 449–456. | Frames attribution as a longstanding measurement problem renewed by increasingly granular customer-path data and a growing number of touchpoints.
• Kelly, J., Vaver, J., & Koehler, J. (2018). “A Causal Framework for Digital Attribution.” Google Research. | Distinguishes assigning credit for an observed outcome from the harder problem of determining whether an advertising exposure actually contributed to causing it.
• Buhalis, D., & Volchek, K. (2021). “Bridging Marketing Theory and Big Data Analytics: The Taxonomy of Marketing Attribution.” International Journal of Information Management, 56, 102253. | Shows how richer digital traces, big-data analytics, customer interactions, and increasingly sophisticated attribution models have expanded the measurement problem rather than simply solved it.
• Gordon, B. R., Moakler, R., & Zettelmeyer, F. (2026). “Predicted Incrementality by Experimentation (PIE) for Ad Measurement.” NBER Working Paper No. 35044. | Contrasts attribution counts and last-click measures with randomized experimentation designed to estimate incrementality — whether advertising caused behavior that otherwise would not have occurred.