Audience measurement answers a different question from everything else here — how many, who, and for how long — and is routinely confused with spending data.
Broadcasters, streamers, agencies and advertisers who need a common currency for a transaction. Ratings exist so a buyer and a seller can agree on what was delivered. Every other source on this site describes money or activity; these four describe people.
Nielsen Audio is the radio currency. Comscore measures television and digital audiences. Edison Research conducts the audio behavior studies the industry quotes. Triton Digital measures streaming audio and podcast delivery.
Nielsen Audio produces market-level radio ratings, using portable meters in the largest markets and diary-based measurement elsewhere. Comscore produces television and cross-platform audience estimates, including local television, and digital audience measurement. Edison Research runs the long-running Infinite Dial study on American audio and digital media behavior and the quarterly Share of Ear study on how listening time is divided across sources. Triton Digital measures streaming and podcast audiences through server-side delivery data.
Together they answer reach, frequency, composition and time spent — none of which any spending source can tell you.
Four different methods, and the differences matter more than the brand names. Panels of metered participants carrying encoded-audio meters. Written or electronic diaries kept by recruited households. Server logs recording actual streaming and download requests. And probability-based telephone and online surveys, which is Edison's method for the Infinite Dial and Share of Ear.
A panel estimates from a sample. A server log counts events. A survey measures stated behavior. These produce genuinely different kinds of numbers and should not be compared as if they were the same currency.
Metered markets report monthly and quarterly. Diary markets report on a much slower schedule — historically as few as one or two measurement periods a year, which means a diary market's book can be many months old when it is being sold against. Streaming and podcast server data is close to real time. Edison's Share of Ear is quarterly; the Infinite Dial is annual and usually released in the spring.
The most common error is treating an old diary book as a current fact. Check the survey period before quoting anything.
Because the two axes get conflated constantly. A client who says radio is not working is usually making a spending or attribution claim, not an audience claim, and answering it with a ratings book misses. Conversely, a spend estimate cannot tell you whether anyone heard the ad.
Edison's work is particularly useful in conversations where you are not the interested party — independent research on how Americans divide their listening time is a stronger argument than anything published by a broadcaster.
Panel-based measurement recruits a sample designed to mirror the market's demographics, records exposure, weights each participant to represent a population segment, and projects. Diary measurement asks recruited respondents to record listening in writing, then applies the same weighting logic to a smaller and more error-prone sample. Server-side measurement counts requests against a technical standard for what constitutes a listen or a download. Survey research draws a probability sample and reports stated behavior with a published margin of error.
Every one of these carries a sampling error that grows as the geography shrinks and the demographic narrows. A ratings number for adults 25-54 in a large market is a fairly stable estimate; the same cell in a small market can swing wildly between books on sample alone.