Measurement methodology: estimated vs measured

Why observations and estimates are kept separate

A PSA campaign report has two kinds of numbers in it. Some record something that happened: a spot aired on a station at a certain time, or a visitor completed a donation. Others are calculations built on assumptions, such as how many people probably saw the spot or what the airtime was probably worth. Both are useful, but they’re different things, and a report that blurs them is claiming more than it can support.

We apply one rule across our datasets, dashboards, exports and written reports:

Never present a modeled number as if it were the station’s official rate. Store the evidence, calculate the estimate, show the source date and label the result.

In practice, we store the evidence behind an estimate separately from the estimate itself. A rate observation is kept with its source, date, daypart and spot length. An audience figure is kept with its survey period, grain and demographic. We calculate the estimate from that evidence, and the report labels the result as estimated and shows the evidence date. When new evidence comes in, we recalculate and keep the old version on file.

Where the data comes from and how it is verified

We don’t rely on any single source by itself. Each estimate draws on several kinds of evidence, and we check them against each other before a figure goes into a report.

• Real transaction invoices. Prices actually paid for airtime on individual stations are one of the rate inputs, alongside published market rate information that we use for context and cross-checks.
• U.S. Census Bureau. American Community Survey population and household universes by age, sex and household, and TIGER/Line geography for station coverage areas.
• TVB (Television Bureau of Advertising). Public television usage levels by daypart and season, used to turn audience shares into ratings and to bring older audience evidence up to the campaign period.
• Radio industry public sources. Published station audience figures, such as those reported by RadioInsight, as anchors for radio estimates.
• Published federal PSA campaign reports, such as NHTSA’s traffic safety campaigns, which report delivery for real public service announcement campaigns and give us an independent back-test.

We verify at three levels. Cross-checks: we compare rates from real transactions with published rates by station and daypart, and investigate any implied CPM far outside the normal range for a market before release. Calibration: we test modeled audience and value against the delivery reported by real public service announcement campaigns. Traceability: each published figure traces to a dated source and a recorded method, and a figure with no traceable input isn’t shown. We use source figures to build our own estimates and credit them. We don’t republish them as tables.

If you get in touch, we’re happy to walk you through the rate evidence and validation behind a specific campaign.

Audit-ready documentation

Any figure in a PSAReach report can be traced to its evidence: the airing record behind it, the rate and audience inputs used, their dates and the method. On request, we’ll prepare a documentation packet your finance team and auditor can review.

Whether and how to record donated media in your financial statements is a decision for your organization and its auditor. We supply the evidence behind each value. We don’t give an accounting opinion.

What is measured

These figures are observations. We state them plainly, without the word “estimated”, subject to any coverage caveat noted once on the report.

• Airings. Each verified occurrence of a creative on a station, with date, local time and creative identifier, from measured occurrence data. The monitoring service’s coverage is stated. A gap in the data doesn’t prove that a station didn’t air the spot.
• Stations, markets and dayparts. Derived from airings by normalizing station identity, assigning the DMA or radio metro and converting the timestamp to a local daypart.
• Creative and spot-length breakdown. From the identifier matched on each airing.
• Conversions. Donations, sign-ups, applications and similar actions recorded in the organization’s own website analytics and Google Ads account.
• Digital reach and frequency. When a paid digital campaign runs on a major advertising platform, we treat the platform’s own reach and frequency reporting as measured and name the platform as the source.

What is estimated and how

Each figure in this section is modeled. Wherever it appears, it carries the word “estimated”, a one-line method and the source period of the underlying evidence.

Commercial-equivalent rate

For a station, daypart and spot length with enough recent observations, the typical rate is a robust central estimate, usually a weighted median or a trimmed estimate that a single unusual buy can’t pull around. Recent, commercially comparable observations are weighted more heavily than older or unusual ones.

Where there are too few observations, the estimate falls back to a published rate card or a market benchmark, and the report says which. A rate-card figure is labeled “rate-card value”, never “commercial-equivalent value”.

CPM

CPM expresses value per thousand impressions. It’s the scale on which broadcast and digital delivery can be compared.

CPM = (Estimated Commercial Value ÷ Estimated Impressions) × 1,000

For donated broadcast, both inputs are estimates, so the CPM is estimated. We also use the implied CPM as a sanity check. A figure far outside the normal range for a market and daypart usually means a rate has been attached to the wrong daypart or an audience figure to the wrong demographic.

Impressions from CPM

When there’s no direct audience estimate but there is a well-supported value and a market CPM benchmark, impressions can be derived. We label the result a CPM-derived estimate, because it stacks one estimate on another.

Estimated Impressions = (Estimated Commercial Value ÷ Estimated CPM) × 1,000

Our main method for impressions is direct: a program or daypart rating multiplied by the market universe estimate for television, and the average quarter-hour audience for radio. For each impression figure, the report states the grain (program or daypart) and the demographic (households or a target group).

CPP and rating points

Cost per point is the price of one rating point in a market. Where no published rating is available, a unit rate and a market CPP benchmark can yield an estimated rating.

Rating Points = Unit Cost ÷ CPP

A rating derived this way is labeled as derived and treated as weaker evidence than a published rating.

GRP

Gross rating points add up the rating points across all the airings in a schedule.

Estimated GRP = (Gross Estimated Impressions ÷ Relevant Audience Universe) × 100

When the impressions are modeled rather than measured by an audience vendor, the result is labeled “estimated GRP”. GRP is a gross exposure measure, so it can be summed across weeks, stations and markets. It doesn’t tell you how many unique people were exposed, and unique reach can’t be derived accurately from gross impressions alone.

TRP

Target rating points use the same arithmetic as GRP, applied to a defined target audience such as adults 25 to 54. TRP needs target-level audience estimates and a target-level universe. Where those are thin, especially in low-audience dayparts, household GRP is the more defensible figure and the report explains why.

Reach and frequency

True reach is unduplicated audience, and it requires data on how much the audiences of different airings overlap. Broadcast PSA campaigns rarely have that data, so reach is modeled. We model it within each market and report it with an average frequency, plus the share reached three or more times.

We state the limits on each reach figure. The model runs on impressions that are themselves estimated, so uncertainty compounds. Reach is reported per market and per period. Market-level reach can be summed to a national figure because DMAs are disjoint populations. Reach within one market is never summed across weeks. Reach across television, radio and digital isn’t combined, because cross-media duplication can’t be measured without single-source data.

How estimates are labeled

A label has three parts, and they always appear together.

1. The word estimated in the figure’s name, on chart axes, in table headers and in exports.
2. The method in one line, such as “program-level household rating times market universe”.
3. The source period of the audience evidence and the compile date of the rate evidence.

Dashboards show an “estimated” badge beside each modeled figure. Written reports put the word in the sentence rather than in a footnote. We compare periods only when the method and source period are the same, and the report says so when they aren’t.

Spot length

Spot length changes value but not audience. A :15 and a :30 in the same break reach the same people, so their estimated impressions are identical. Their commercial-equivalent value differs, because a :15 typically prices at roughly half to 60 percent of a :30 and a :60 at more than a :30 but less than double.

We carry :15, :30 and :60 airings separately and never normalize a schedule to “:30 equivalents”. Length is applied to the rate and to the value calculation. It’s never used to scale impressions, GRP, reach or frequency.

What PSAReach will not do

• Publish modeled station prices at scale as if they were official. Rate estimates are labeled estimated, and we don’t publish station rate tables.
• Claim measured reach for broadcast. Broadcast reach is modeled and is always labeled estimated.
• Sum reach across periods. Reach in the same market is reported per period. National totals sum across markets only.
• Claim combined cross-media reach. Television, radio and digital reach are reported separately. Where paid digital is targeted to fill gaps that broadcast left, the report describes the overlap as small by design, not as measured.
• Present a rate-card total as market value. Rate-card figures are labeled as list prices.
• Scale audience by spot length. Length affects value only.

Holding to these rules makes reports slower to produce, and it means each figure can stand up when someone asks where it came from. The glossary defines each term used here, the calculators apply the same formulas, and the Media Intelligence page describes what our audience, reach and value estimates cover.