Arthur Nielsen had conducted successful analysis of radio programming in the 1930s and 1940s. Before that, Nielsen worked at testing new products and measuring sales figures to determine what kinds of people bought certain types of products He developed the TV ratings system in 1950. In 1954, he started using the Recordimeter on TV sets in sample homes and noted what they watched. Nielsen fine-tuned his system until his death in 1980.
Nielsen Media Research places recording devices in 25,000 U.S. households that represent a broad spectrum of ages, races and viewing tastes. The devices monitor the time the TV is on, and the station watched, to assemble the ratings data. Nielsen households cannot divulge to others that they have the boxes. The Nielsen households are paid a small amount, up to $40 a month in some cases, and can act as a Nielsen family for up to five years.
Nielsen data is divided into ratings and shares. The ratings are a percentage of everyone with a television set watching a show, regardless of whether the set is on or not. The shares are a percentage of everyone with their televisions on watching a show during a given time. For example, if there are 100 million households with TVs, and 20 million of them watch a given cop show every Thursday at 8 p.m., it would have a rating of 20 (or 20 percent). If only 50 million households have the TV turned on Thursday at 8 -- with 30 million watching something other than the cop show -- it would have a share of 40, or 40 percent of the televisions being watched at the time.
Because the Nielsen ratings sample only a small percentage of the viewing public, the data is not comprehensive. Furthermore, the Nielsen system was developed when there were only three major networks. As of 2011, viewers have access to hundreds of stations through cable and satellite services, not to mention DVRs, video on demand, and online viewing services like Hulu. Critics, such as Aresh Amel of "Screen Digest," contend that Nielsen struggled to adapt to these changes. That means the ratings do not accurately reflect how many people are actually watching a given show.