Sports TV ratings methodology

What this site is

A reference record of publicly reported audience figures for US sports telecasts. We collect every figure that a measurement company, network, platform, league, or credible outlet publishes; we record it with its source, its exact metric definition, its publication date, and which provider the figure is attributed to; we compute the spreads, revisions, and standings that follow from the record. Measurement companies on the board: Nielsen, Comscore, iSpot, VideoAmp, and AdImpact, plus platform and league self-reports.

What we do not do

We do not adjudicate. A spread documents that published figures disagree; it does not assert which figure is correct. We license no data from any measurement provider and are a client of none. We accept no sponsorship from any measurement company. We publish no opinion content.

Which figure leads

Where a telecast page shows one figure first, it is the Nielsen average audience for the full telecast. Nielsen is the currency the US television market transacts on, so its figure is the one most other published figures are compared against; leading with it is a transactional convention, not a judgment that it is more accurate than any other provider's figure. Every provider figure on the page has equal standing in the published record, and every disagreement between providers is computed from the figures as published and never resolved.

Where figures come from

Every figure on this site was voluntarily published to the public record (press releases, company sites, trade coverage), and every figure links to its source. A trade article naming a provider is not confirmation: we reconcile each figure to that provider's own release before treating the attribution as confirmed. Press reports often conflate AdImpact and iSpot; when we cannot tell which provider produced a number, we withhold it rather than guess. Figures we cannot fully verify against an open source are marked partial confidence. We never reproduce licensed datasets; facts, individually reported with attribution, are what the Ledger contains. When we fetch sources we identify ourselves; we do not impersonate a browser.

Metric definitionswhy two true numbers differ

Average audience

mean viewers across the measured span

Peak audience

highest measured block (usually 15 minutes)

Total reach

unique viewers who watched any qualifying portion; reach typically runs well above average and the two are not comparable

Cuts

the same broadcast can be measured as the full telecast, the match only, or a stated window; combined multi-network totals are marked composite, with components and combination rule shown where published

Platform scope

where a source says which platforms an average covers, the figure is marked TV only, TV + streaming, or streaming only; where the source does not say, nothing is marked. A TV-only figure and a TV + streaming figure for the same telecast are shown as two definitions of the audience, never as a disagreement between them

Self-reported

figures a platform or league publishes about its own audience, with no independent measurement named, carry this flag

Split window

a national broadcast window that carried different games in different markets. The published figure measures the window, not the single game, so it is marked split window and the source's own coverage wording is quoted beside it ("mostly", "in most markets", "in 78% of markets"). A figure without that mark measures the telecast itself. A split-window figure whose qualifier is missing is not published

Measurement eras

Methodologies change: households → viewers, out-of-home added (2020), Nielsen Big Data + Panel (Sept 2025), and challenger certifications (2024– ). Figures from different eras are different instruments. Era boundaries are marked on every comparison surface, and cross-era rankings are flagged rather than silently merged. See the Era Guide.

A figure also carries a measurement method, which instrument produced that number, separate from the era calendar. When Nielsen publishes both Big Data + Panel and panel-only for the same telecast, both rows stay live. That pair is not a currency disagreement and is not collapsed into a revision. When two eras overlap on a date, we leave the era unmarked rather than guess; the method label still identifies the instrument.

Statuses and revisions

Figures are tagged final, preliminary, revised, or partial confidence. When a published figure changes, we do not overwrite it: the original and the revision both remain, with dates, forming the revision record.

Ledger columns

Telecast

the telecast the figure is attached to, with its air date. A figure with no telecast row is listed without one.

Measured by

the measurement company, or the platform or league that published its own figure. "Blend" marks a published figure that mixes providers.

Metric

average audience, peak audience, or total reach, as defined under Metric definitions.

Cut / basis

the network shape of the figure (full telecast, match only, stated window, combined, per network), the span that was averaged, the platform scope (TV only, TV + streaming, streaming only) where the source states it, and the measurement method (Big Data + Panel or panel-only) where the provider states it.

Value

viewers, as published. Stored as a whole number; we do not round.

Type

currency: a figure from an independent measurement company. Self-report: a platform or league reporting its own audience. Combined definition: a published total that sums components.

Status

final, preliminary, revised, or partial confidence, as defined under Statuses and revisions. Superseded marks a figure that a later revision replaced; both stay listed.

Attr

provider attribution. Confirmed: reconciled to the provider's own release. Press-attributed: a reporting article names the provider, corroborated by a second source before publishing. Ambiguous attributions are never published.

Published

the date the figure entered the public record. For figures carried from a history table, it is the date we recorded it.

Source

the page the figure was taken from. Every figure links to its source. A press-attributed figure carries that mark beside its source; an unmarked source is confirmed against the provider's own release (see Attr).

Publisher

who published a self-reported figure.

Geo

the geography the figure covers.

Era

a span during which one measurement methodology was in effect, as described under Measurement eras.

Date

the telecast's air date. On a series row, the year.

Figure

the telecast's lead figure: Nielsen average audience, full telecast, as published. Blank when Nielsen has not published one.

Name

the league, network, or team; a team page covers the franchise under every name it has carried.

Telecasts

how many telecasts with a published figure the entity has.

Aired

the telecast's air date.

First figure

the date the first public figure for the telecast appeared.

Lag

whole days from air to first figure. The Release Calendar lists only first figures published within 60 days of air; a figure carried from a history table later than that is an archive entry, not a release.

Coverage honesty

Not every telecast receives a public figure from every source. "Not published" is recorded, and coverage rate (how often each source publishes publicly) is itself a tracked statistic. This is a record of what the industry chose to publish, stated as such.

Some telecasts, most historical ones, have every figure traced to a single source page, with no second source published to compare against. Those pages say so. Single-source is not corroborated, and we do not imply otherwise; the historical Super Bowl series (1967 onward) is the largest such run, carried from one Nielsen-sourced history table.

Chart eligibility

Rankings rank one metric: average audience, currency-of-record, per telecast. A preliminary figure charts until a final or revised figure for the same telecast replaces it (ADR-120); its status is shown on the telecast page. Composites, reach figures and self-reported figures appear on telecast pages but do not chart. Cross-era chart records are marked. Eligibility disputes are documented on the relevant telecast page, not silently resolved.

Corrections

Extraction errors, once found, are corrected within one publication cycle and logged publicly on the corrections page with before/after values. Report an error: every page carries a report link tied to its claim ID. A report is checked against the source it cites by the pipeline: the cited page is re-fetched and re-read under the same rules as any other source, and the record changes only if that reading does. Nothing is edited by hand, and a report with no public URL closes without a change.

Self-reported submissions

Networks, leagues, teams, publishers, and measurement providers may point the site at a public page where an audience figure is stated. The page is fetched and read under the same rules as every other source; the metric, provider, and telecast are taken from the page, never from the submission, and nothing publishes until the pipeline confirms the figure there. A figure with no public source is not recorded. A licensed feed or private export is not accepted.

Reuse and attribution

The RatingsReference compilation — its selection of figures, the claim-level structure, page templates, derived statistics, charts, and the JSON record behind every page — is published under the Creative Commons Attribution 4.0 license (CC BY 4.0). You may copy, redistribute, and adapt it for any purpose, including commercial use, provided you credit RatingsReference.com and link to the record you used. Individual audience figures are facts reported as published by the source named on each claim; the license covers our compilation of them, not the figures themselves, and each figure's own source should be cited alongside ours. Every JSON record carries the license URL and this scope statement.

Automation disclosure

This site is compiled and maintained by an automated pipeline. Every figure passes independent double extraction and confidence gates before publication; figures failing those gates are withheld, not guessed. The methodology above is enforced in the pipeline's code, and this page changes only when the pipeline does.