Every week, headlines declare that a show is a “ratings hit” or a “flop”. A series is cancelled after three episodes, or renewed for a fifth season nobody expected. Behind those decisions sits a measurement system most viewers never see — and streaming has changed what the numbers mean. Here is how television ratings actually work, who counts them, and how to read them without being misled.
What a TV rating actually is
A television rating is an estimate, not a headcount. Nobody counts every single viewer. Instead, a measurement company studies a sample of households and projects what that sample watched onto the whole population.
The two numbers you will see most often are the rating and the share, usually written together like “3.4/8”. The rating is the percentage of all television households in the measured population who watched a programme. The share is the percentage of households that actually had their television switched on at that time who were watching that programme. So a show with a 6.0 rating and a 15 share was watched in 6 per cent of all TV homes, and among the homes with the TV on, 15 per cent chose it. Share is the competitive measure — it tells you who won the time slot — while the rating measures absolute reach.
Ratings are also broken down by age group, and the one the industry watches most closely in the United States is adults aged 18 to 49. Advertisers pay a premium to reach that group, so a show can have a modest total audience and still survive if its 18-to-49 number is strong.
How Nielsen measured television in the old days
Nielsen is the company whose name became a byword for ratings. The firm was founded in 1923 by market research pioneer Arthur Nielsen, and the modern ratings era began in 1942, when Nielsen introduced the Audimeter, a device attached to a radio — and later a television — that recorded when the set was on and which station it was tuned to.
For decades, the system relied on a recruited panel of households, selected through probability sampling and weighted against census benchmarks so that it mirrored the country. In the 1980s, the People Meter arrived: a box on top of the television with buttons each household member pressed when they started and stopped watching, which added demographic detail — who was watching, not just that a set was on. Paper diaries were used for years in smaller markets, especially during the “sweeps” months of February, May, July and November.
It was an ingenious system, and it was also widely criticised. A panel of tens of thousands of homes was standing in for more than 100 million, and viewing on phones, tablets and streaming apps was hard to capture at all.
What Nielsen does now: Big Data + Panel
In September 2025, Nielsen moved its national television measurement to a new methodology called Big Data + Panel, which it describes as its most accurate measurement to date. The idea is to pair the two things the old system kept apart: the scale of big data and the demographic accuracy of a panel.
The big data comes from cable and satellite set-top boxes and from automatic content recognition in internet-connected smart TVs — Nielsen says it covers 45 million households and 75 million devices in the United States. That data shows what is playing on a screen, but only at device level: it cannot tell who in the household is watching. That is where the panel comes in. Nielsen’s panel still consists of more than 42,000 homes and around 100,000 people, and it is used as the source of truth to calibrate the big data — assigning viewing to the right individuals and projecting audience estimates to the whole TV population. Nielsen says the combination sharply reduces zero ratings for smaller programmes.
There are two practical consequences for anyone reading ratings news. First, numbers measured under the new system are not directly comparable with the old panel-only numbers — a show’s audience can look different simply because the ruler changed. Second, Big Data + Panel figures generally take two to three days to arrive, so the instant “overnight” numbers quoted the morning after a broadcast are preliminary; under Nielsen’s press policy, public claims by broadcasters, networks and streamers should be based on Big Data + Panel data, with an exception for single-day events.
The Gauge and the Streaming Top 10
Nielsen also publishes two products that have become the standard references for the streaming era. The Gauge, released monthly, shows the share of total television time going to broadcast, cable and streaming. Every week, Nielsen publishes its Streaming Top 10, a ranking of the most-watched streaming programmes and films in the United States, measured by total minutes viewed by people aged two and older. Because it measures minutes rather than a streamer’s own definition of a “view”, it is the closest thing to an independent yardstick for streaming.
How streamers count a “view”
Streaming platforms do not use Nielsen’s method for their own announcements — they count from their own servers, and each one defines a “view” differently. Netflix, which reports more viewing data than any other streamer, explains its method publicly. Its weekly Top 10 lists, measured Monday to Sunday and published on Tuesday, are ranked by “views”, calculated by dividing the total hours viewed by a title’s runtime. The change was made in June 2023, replacing a ranking based on total hours viewed, because long titles had an unfair advantage. Netflix’s all-time most popular lists count views from the first 91 days after release.
That formula matters, because a Netflix “view” is not a count of distinct people. Ten million hours watched of a two-hour film equals five million views — but those hours could come from five million people watching once, or one million people watching five times. Hours viewed is watch time, not a headcount, and neither measure reveals completion rates, subscriber growth or profitability. Other streamers, including Disney+ and Hulu, use similar hours-divided-by-runtime calculations, but most publish far less data than Netflix, and some announce a “views” figure without explaining the definition at all. When a platform reports its own number, treat it as a self-reported claim, not an independent measurement — and always check whether an independent source such as Nielsen’s Streaming Top 10 tells the same story.
Why a show gets cancelled or renewed
Ratings are only one input into the decision, and in streaming they are not even the input the public can see. For a traditional broadcast or cable network, the maths is relatively direct: a show earns money by selling advertising against its audience, and the price of that advertising is set by the ratings, above all in the 18-to-49 demographic. A show with falling ratings earns less from advertising and gives a weak lead-in to the programme that follows it. High production costs, an expensive cast, or a network that does not own the show outright can all tip a borderline series towards cancellation.
Streaming changed the equation. A streaming service earns money from subscriptions, not from selling audiences to advertisers — although advertising-supported tiers now exist on most major platforms. What matters is whether a show brings in new subscribers, keeps existing ones from cancelling, and justifies its cost relative to its audience. A niche series with a devoted audience can be worth more to a streamer than a broadly watched but unremarkable one, because devoted viewers are less likely to cancel. Cost still matters enormously: a spectacularly expensive drama needs a much larger audience to justify another season than a cheap reality format does.
That is why the same audience can mean renewal on one service and cancellation on another, and why a show that “everyone is talking about” can still disappear. Talk is not measured; viewing, retention and cost are.
How to read a ratings headline
A few habits will keep you from being misled. First, ask who measured it: a Nielsen figure and a platform’s own press release are not the same kind of evidence. Second, ask what the number counts: a rating, a share, total viewers, minutes viewed and “views” are five different things, and a “view” on one platform may not equal a “view” on another. Third, check the time window: live viewing, viewing within seven days and viewing within 35 days can produce very different numbers for the same episode, and streaming totals grow for weeks after release. Fourth, check the demographic: a headline about total viewers and a headline about adults 18 to 49 can describe the same show in very different terms. And finally, remember that rankings are relative — a show can top a weekly chart in a quiet week with an audience that would have placed it fifth a month earlier.
Beyond the United States
In the United Kingdom, television audiences are measured by Barb Audiences Ltd (formerly the Broadcasters’ Audience Research Board), founded in 1981 and jointly owned by the BBC, ITV, Channel 4, Channel 5, Sky and the advertising industry body the IPA. Barb combines a panel of around 7,000 UK homes — roughly 16,000 people — with device-level census data from streaming services, fused through a process called Dovetail Fusion, and publishes overnight ratings at about 9.30 a.m. the following morning. In Australia, OzTAM is the official source of television audience measurement, with Nielsen as its research supplier operating the panels. Each country has its own measurement body, so a “ratings hit” in one market says nothing about another.
The bottom line
TV ratings are not a simple count of who watched what. They are an estimate built from a sample, calibrated against census data, and — since 2025 in the United States — fused with data from tens of millions of devices. Streaming added a second system, in which platforms report their own figures using definitions they choose themselves. Knowing which system produced a number, what it counts, and over what period is the difference between understanding a hit and being fooled by one.