When TV viewers or entertainment professionals in the United States mention "ratings" they are often referring to Nielsen Ratings, a system developed by Nielsen Media Research to determine the audience size and composition of television programming. Nielsen Ratings are offered in over forty countries.
Other television audience measurement systems are available from other companies, as well as systems developed in joint ventures with Nielsen Media Research, such as AGB Nielsen Metivodia Research.
Nielsen's ratings calculation, also called Cume Rating (or "Reach"), measures the number of unique viewers or households tuned to a television program in a particular time period during a week. The Cume itself is calculated by dividing the number of unique viewers or households by the total number of estimated available households/viewers/listeners possible. This gives a percentage Cume (cumulative) rating.
The system has been updated and modified extensively since it was developed in the early 1940s by Arthur Nielsen, and has since been the primary source of audience measurement information in the television industry around the world. Since television as a business makes money by selling audiences to advertisers, the Nielsen Television Ratings are the single most important element in determining advertising rates, schedules, and program content.
The company is part of the Nielsen Company, formerly known as VNU and owned by a consortium of private equity firms including Blackstone Group, KKR and Carlyle Group. Its production operations are located in its Brooker Creek Global Technology and Information Center in Oldsmar, Florida.

Measuring ratings
Nielsen Television Ratings are reported by ranking the percentage for each show of all viewers watching television at a given time. As of September 2006, there are an estimated 115.2 million television households in the USA. A single national ratings point represents 1%, or 1,152,000 households for the 2006-07 season. Share is the percentage of television sets in use tuned to a specific program. These numbers are usually reported as (ratings points/share). For example, Nielsen may report a show as receiving a 9.2/15 during its broadcast, meaning 9.2%, or 10,598,400 households on average were tuned in at any given moment. Additionally, 15% of all televisions in use at the time were tuned into this program. Nielsen re-estimates the number of households each August for the upcoming television season.
Nielsen Media Research also provides statistics on estimated total number of viewers, and on specific demographics. Advertising rates are influenced not only by the total number of viewers, but also by particular demographics, such as age, sex, economic class, and area. Younger viewers are considered more attractive for many products, whereas in some cases older and wealthier audiences are desired, or female audiences are desired over males. Television ratings are not an exact science, but they are a powerful force in determining the programming in an industry where millions of dollars are at stake every day.
Because ratings are based on samples, it is possible for shows to get 0.0 rating, despite having an audience; CNBC talk show McEnroe was one notable example.

Ratings / share and total viewers
Nielsen provides viewership data calculated as the average viewership for only the commercial time within the program. This "Commercial Ratings" first became available on May 31, 2007. Additionally, Nielsen provides different "streams" of this data in order to take into consideration delayed viewing (DVR) data. Nielsen's clients now have access to all the data they need to develop individualized minute-by-minute ratings of national commercials by demographic group for all national television programs, including DVR playback at any interval up to seven days.

Commercial Ratings
Much of the ratings system, however, still consists of the completion by viewers of ratings diaries, in which a viewer records his or her viewing habits, generally for a week, in exchange for being advanced a nominal amount ($30 in the United States.) These diaries play an especially important role during the four sweeps periods conducted in February, May, July and November in an attempt to measure smaller local market audiences in markets that are not covered by People Meter samples already. (Other, smaller sweeps are conducted through the year in the markets large enough to be measured by non-demographic meters, but not large enough to be measured by the demographic meters (people meters.)
The term "sweeps" has two meanings. One refers to how the diaries were handled by Nielsen Media when the ratings were first produced: They are mailed to the households and processed by starting on the East Coast and "sweeping" across the nation. The other refers to television programming during the months of November, July, and/or May, in which eagerly anticipated programs are deliberately scheduled in order to boost television ratings.
Television networks and other programmers make unusual efforts to attract additional viewers during these periods, including airing mostly first-run programming as opposed to repeats, airing more special broadcasts, and including special content in programming such as guest stars, controversial and unexpected plots or topics, extended episodes, finales, and increased competition in advertising. Even news programs are often involved, airing especially controversial or titillating investigative reports and promotions. For this reason, the "sweeps" system of national ratings has been criticized as not representative of typical programming, and encouraging an increase in content of concern such as violence and explicit sexuality. Outside of these peak periods it is more common to see reruns of television programs.
Nielsen 2007 Sweeps Dates:

Annual top-rated shows

List of most-watched television episodes
List of television stations in North America by media market
BARB