Nielsen Media Research

• 1. Introduction

  ›

  Nielsen Media Research holds a quiet but decisive power over American television. Every autumn, when networks decide which shows live and which die, the verdicts trace back to a single question: how many households were watching? The answer comes from Nielsen, a firm whose methods have shaped prime-time schedules since the 1950-51 television season. How does a company turn the scattered habits of millions of viewers into a number that determines whether a show gets renewed or cancelled? And what happens when those methods, built for a world of living-room sets and paper diaries, collide with streaming, smartphones, and the fractured viewing landscape of the 21st century? Those are the questions this documentary sets out to answer.

• 2. The Roots of Measurement

  ›

  Arthur C. Nielsen began his career in the 1920s doing marketing research and performance analysis. His firm expanded into radio market analysis in the late 1930s, and by 1942 it had produced the Nielsen Radio Index, designed to give broadcasters hard statistics about who was listening to which shows. The first Nielsen ratings for radio programs were released in the first week of December 1947. They tracked the top 20 programs across four categories: total audience, average audience, cumulative audience, and homes per dollar spent on time and talent.

  Radio was the laboratory. When television arrived, Nielsen moved the same methodology onto the new medium. By the 1950-51 season, the Nielsen Television Index was up and running. At its core was the Audimeter, a device that attached directly to a television set and recorded which channels were viewed onto a 16mm film cartridge. That cartridge was then mailed each week to company headquarters in Evanston, Illinois. To persuade households to accept the machine, Nielsen offered a practical incentive: free television repair service, provided by TV Index representatives, at a time when vacuum tube sets broke down often and repair costs were real.

  The Audimeter traced its lineage to an earlier device built for the 1942 Radio Index. The logic was the same: let a machine record what people actually watched, rather than trusting them to remember. That instinct toward passive, mechanical measurement would define Nielsen's approach for decades, and the tension between what machines record and what people report would never fully go away.

• 3. Diaries, Sweeps, and the Ritual of February

  ›

  In 1953, Nielsen added a paper diary system alongside the Audimeter, calling it the Nielsen Station Index Service. Households recruited by the company wrote down what they watched and when, allowing Nielsen to gather data on local markets and specific demographic groups that a single national meter could not capture. Interviewers based in Oldsmar, Florida, and Radcliff, Kentucky, coordinated the distribution of these diaries.

  The term "sweeps" dates from 1954. That year, Nielsen collected diaries from households in the Eastern United States first, then worked its way across the country, effectively sweeping west. The four sweeps months settled into a fixed pattern: February, May, July, and November. Seven-day diaries were mailed to a rotating panel of homes each week of the sweeps period; homes with DVRs received eight-day diaries to capture time-shifted viewing. At the end of the month, all the weekly data was pooled.

  Each year until 2018, Nielsen processed roughly two million paper diaries from across the United States. Local television stations relied on sweeps data to set advertising rates and make scheduling decisions. The November, February, and May sweeps were considered the most consequential; the July sweeps could still affect local personnel decisions. One notable disruption came in the 2008-09 season, when the February sweeps period was shifted to March to accommodate the digital television transition, which had been scheduled for the 17th of February 2009. When that transition date was later moved to June 12, Nielsen kept the sweeps in March rather than shifting the schedule a second time.

  By June the 28th of 2018, the paper TV diary rating service was retired entirely, replaced by year-round electronic measurement.

• 4. From Film Cartridges to People Meters

  ›

  The Audimeter of 1950 required households to physically mail film cartridges to Evanston. That lag meant ratings arrived days after a broadcast. In 1971, the Storage Instantaneous Audimeter solved the problem by sending electronically recorded viewing data through a telephone line, making overnight ratings possible for the first time.

  The next leap came in 1987, when Nielsen introduced the People Meter. Unlike the Audimeter, which recorded only that a set was on and tuned to a channel, the People Meter tracked individual viewing habits within a home. Each member of the household logged in separately, and the device transmitted that data nightly via telephone line. Researchers could now study viewing patterns on a minute-by-minute basis, recording exactly when someone changed channels or turned off the set.

  Nielsen later replaced the People Meter with Portable People Meters, which separated household members through individual login credentials and allowed the company to break viewing information into distinct demographic groups. The Local People Meter, introduced in New York and Los Angeles in 2003, brought the same granularity to major local markets rather than relying on a national sample. The LPM shifted measurement from the quarterly sweeps rhythm to a continuous, 365-day cycle. By 2025, LPMs are deployed across the 25 largest Designated Market Areas, while mid-sized markets use a combination of Set Meters and anonymous Return Path Data drawn from cable and satellite set-top boxes.

  In January 2025, Nielsen received accreditation from the Media Rating Council for its Big Data + Panel product. That system combines a traditional panel with data from smart TVs and covers person-level estimates across 45 million households and 75 million devices.

• 5. What a Rating Actually Means

  ›

  A Nielsen rating is a fraction. The denominator is the total number of television-equipped households in the United States; for the 2017-18 season that figure stood at 119.6 million, a number Nielsen re-estimates every August. The rating itself expresses what percentage of all TV households were tuned to a given program. Share, the companion figure, asks a narrower question: of the households that had a television on at that moment, what share was watching this particular show?

  When a network reports a result such as 4.4/8, it means an estimated 4.4% of all TV-equipped households were tuned in, while 8% of households actually watching television at that hour were on that channel. Because the system relies on samples, a show can mathematically record a 0.0 rating even with a real audience. CNBC's talk show McEnroe achieved exactly that. So did The CW's show CW Now, which recorded two 0.0 ratings in the same season.

  The rating and share figures, however, are not the primary language that advertisers speak. Specific demographic groups drive ad rates far more than raw household counts. According to Advertising Age, during the 2007-08 season, ABC charged $419,000 per commercial during Grey's Anatomy, compared to $248,000 for a spot during CBS's CSI: Crime Scene Investigation, even though CSI drew nearly five million more total viewers. The 18-49 demographic, where Grey's Anatomy outperformed CSI, explained the gap. NBC charged almost three times as much for a commercial during Friends as CBS charged for Murder, She Wrote, despite the two series drawing comparable total audiences during the seasons they aired simultaneously. During the 2009-10 season, Glee on Fox and The Office on NBC each drew fewer total viewers than NCIS on CBS; their average commercial rates were $272,694 and $213,617 respectively, compared to $150,708 for NCIS.

• 6. Controversy and the Limits of the Sample

  ›

  A sample is only as good as its representativeness, and critics have challenged Nielsen's sample on multiple fronts. In many local markets during the 1990s, the statistical difference between a rating that renewed a show and one that cancelled it was smaller than the margin of error in the sample itself. The show with the marginally higher number survived; the show with the marginally lower number did not.

  In 2004, News Corporation retained the public relations firm Glover Park to run a campaign arguing that Nielsen's People Meter system underreported minority viewing. The campaign's advocates claimed this created a de facto employment bias against minority actors and writers. Nielsen responded by releasing its sample composition data. That data showed African American households represented 6.7% of the Nielsen sample against 6.0% in the general population, and Latino households represented 5.7% against 5.0%. By October 2006, News Corporation and Nielsen reached a settlement, with Nielsen agreeing to spend an additional $50 million to guard against underreporting minority viewing in the new electronic system.

  A different structural problem surfaced as cable expanded the number of viewable networks: the margin of error grew because each individual channel commanded a smaller slice of the sample. Response bias added another layer. Households that know they are part of the Nielsen sample sometimes alter their viewing or reporting habits; self-reported diary counts have at times run higher than counts gathered by electronic meters. In 2009, Nielsen lost the contract for measuring Irish advertising, and agencies described the resulting data as "disastrous" and too inaccurate for clients to use.

  Nielsen lost accreditation from the Media Rating Council in 2022 due to inaccurate data reporting during the COVID-19 pandemic, then regained it in April 2023.

• 7. Streaming and the Fractured Count

  ›

  By September 2020, Nielsen had begun compiling a weekly top 10 list of most-watched shows on streaming platforms. That list arrived two decades after the viewing landscape had started fragmenting. In 2005, Nielsen began measuring DVR usage, and initial data confirmed that time-shifted viewing would alter ratings in ways advertisers were reluctant to accept. A year after those results appeared, networks were still not factoring time-shifted viewing into their advertising rates.

  In July 2017, Nielsen announced it would include select programs from Hulu and YouTube TV in its Digital in TV Ratings system. By around October 2017, it had begun tracking select programs from Netflix as well. The mechanism relies on a tag that partnering distributors embed in the program; Nielsen then tracks the tag through its meters. Distributors retain the right to decide whether those ratings can be made public.

  Yet even as the tracking expanded, a counting problem persisted. As of 2013, internet streams were excluded from standard ratings if the program carried no advertisements or ran different advertising from the broadcast version. In 2012, there was no way for NBC to determine whether any of the roughly 111.3 million traditional viewers of Super Bowl XLVII overlapped with the 2.1 million who watched the live stream, because the two figures were counted through separate systems.

  In April 2015, Nielsen executive vice president of global product leadership Megan Clarken stated at a Coalition for Innovative Media Measurement summit that Nielsen could count digital viewers in its reports, but was prevented from doing so by a set of rules devised by networks and advertising industries, last revised in 2006. Under those rules, Nielsen could only count viewership for television-originated broadcasts and had to exclude digital platform viewers when the advertising load differed or when no linear watermark was present. In December 2025, it was announced that Roku's streaming data would be integrated into Nielsen's Big Data + Panel, adding one more source to a measurement system that has been chasing its audience across screens for more than two decades.