The Man Who Invented the Rating: Archibald Crossley and the Birth of Audience Measurement

A researcher, not a broadcaster, invents the audience number

It is worth pausing on who Archibald Crossley actually was, because it is not the answer most people assume. He was not a radio executive trying to prove his own network's reach. He was a market researcher, working in the emerging discipline of opinion and consumer research alongside figures like George Gallup, and he was solving a specific commercial problem: advertisers who were used to buying newspaper space on the strength of an audited circulation number had no equivalent figure for the new medium of radio.

That origin matters for how the resulting method should be read. Crossley's ratings were not built to describe radio for its own sake. They were built to make radio legible to an advertising market that already knew how to buy against a number, which is why the figure he produced was named, from the start, in commercial rather than descriptive terms.

The Cooperative Analysis of Broadcasting, 1930

In 1930, Crossley formed the Cooperative Analysis of Broadcasting, commonly known as CAB, the first subscription national radio ratings service in the United States. The method was straightforward by later standards: researchers placed telephone calls to random households and asked respondents to recall what programs they had listened to, initially the previous day and later the same day, then aggregated those recollections into a national figure.

This is, in the vocabulary later measurement science would use, a telephone-recall methodology, and it is the earliest instrument in the entire lineage this article belongs to. Every subsequent method, Hooper's coincidental calling, Nielsen's mechanical meter, Arbitron's diary, the modern hybrid of panel and device-level data, exists because CAB's recall approach eventually proved insufficient and needed replacing.

Coining the number that outlived the method

Crossley's most durable contribution is not the specific telephone-recall technique, which did not last two decades. It is the word he attached to the resulting figure: "rating." The term, and the closely related concept of "share" that followed it, became the permanent vocabulary of audience measurement, surviving every methodological revolution that came after, mechanical metering in the 1930s and 40s, the diary in the 1950s and 60s, the people meter in 1987, and the passive device-level data of the 2020s.

That persistence is itself worth noting. A "rating" today is produced by an entirely different evidentiary process than the one Crossley invented, yet it is still called by his name. The word survived because it named a commercial function, a single comparable number advertisers could buy against, that has remained constant even as every method for producing it has been replaced.

Why recall lost: the sixteen-year run of CAB

Crossley's method carried an obvious vulnerability: it depended on a respondent correctly remembering what they had listened to, sometimes a full day earlier. Memory is not a neutral instrument, and researchers of the era already understood that recall-based self-report introduces its own systematic bias, distinct from and generally worse than the bias of asking someone in the moment.

That vulnerability is exactly what C. E. Hooper exploited starting in 1934. Working with design input from George Gallup, Hooper developed the "telephone coincidental" method: calling households during a broadcast and asking what they were listening to right then, removing the memory step entirely. Hooper's in-the-moment method was judged more accurate than Crossley's day-after recall, and it progressively displaced CAB in the marketplace. CAB dissolved in 1946, sixteen years after Crossley founded it, as Hooper became the de facto industry standard.

This is the first documented case in the measurement lineage of one method displacing another explicitly on accuracy grounds, and it sets a pattern that repeats for the next ninety years: every new instrument in this history exists because the one before it had a known, articulable flaw.

What Crossley got right, and what he could not have solved

It would be unfair to read Crossley's obsolescence as failure. He solved the actual problem in front of him, which was not "build the most accurate possible audience-measurement instrument." It was "build something advertisers can buy against, for a medium that currently has nothing." A telephone-recall survey, however imperfect, was a working answer to a problem that had no answer at all in 1930, and it created the commercial category, national ratings for a broadcast medium, that every subsequent, more accurate instrument then competed to serve.

What he could not solve, because measurement science had not yet developed the tools to see it, was the specific bias his own method introduced. That diagnosis had to wait for Hooper's coincidental method to demonstrate, by producing different and more trusted numbers, that recall itself was the problem. This is a useful pattern to hold in mind for any measurement claim made today: a working number is not the same as an accurate one, and the surest way to find out which you have is a second, differently-built method to compare it against.

The lineage this founding actually established

Crossley's CAB is the first entry in a documented, continuous ninety-plus-year record of audience-measurement instruments, each invented to fix a specific, named flaw in the one before it: recall gave way to coincidental calling in the 1930s and 40s, coincidental calling gave way to the mechanical Audimeter for national panels, the Audimeter's reach was extended by the far cheaper diary for markets a wired meter could not economically cover, and the diary itself has been progressively displaced by passive, device-level data since the 1987 people meter. No instrument in that chain has ever produced a true census. Every one, including the most sophisticated modern hybrid, is a model built from a sample, with its own documented bias.

Crossley's 1930 method establishes two facts that should govern how anyone reads a media-measurement number today: first, that "rating" is a commercial vocabulary word that has survived a century of completely different underlying methods, so the word alone tells you nothing about how the number beneath it was produced; and second, that every method in this lineage, including whatever produces today's attention or visibility figures, is a model with a discoverable limitation, not a census, and deserves the same scrutiny Hooper applied to Crossley in 1934.