A two-stage audience intelligence methodology that starts with the entire population inside a station's signal footprint and refines it with real listener data. Transparent, reproducible, and built on public federal data.
Census + Cardigan + Format Affinity
Every station, every market. No survey required.
Engagement + In-Market Surveys
Real listener data tunes the foundation.
Traditional audience measurement starts small: recruit a few hundred diary keepers, hope they remember what they listened to, then extrapolate to millions. If the sample is flawed, everything built on it is flawed.
Cardigan inverts this entirely. We start with the biggest possible foundation — the actual population living inside each station's FCC-certified signal contour — and work inward using research-backed demographic filters. Then we refine those estimates with real survey responses and engagement data.
The result: audience intelligence that is transparent at every step, reproducible by anyone with access to the same public data, and continuously improving as we collect more listener feedback.
For every FM station in every market, Cardigan builds a baseline audience estimate from eight analytical steps — using only public federal data and established industry research. No survey sample required. This is the foundation.
We use each station's FCC-certified service contour to define the actual listening area. The 54 dBu contour (the city-grade signal boundary) is preferred; when unavailable, we fall back to the 60 dBu contour. This is the station's real signal footprint — not an arbitrary metro boundary or ZIP code approximation.
We aggregate US Census tract-level demographics for every census tract whose centroid falls within the contour. This gives us the total population that can hear the station, broken down by age, gender, ethnicity, and income. The result: a precise demographic profile of the station's potential audience.
Not everyone in the coverage area listens to radio. We apply age-specific radio usage probabilities derived from Edison Research Infinite Dial data to determine how many people in each age bracket actually tune in during a given week. Adults 35-64 listen at much higher rates than teens; this step accounts for that reality.
Each radio format — Country, AC, CHR, Rock, News/Talk, and dozens more — has a demographic affinity profile based on decades of industry research. We weight the listening population by how likely each demographic segment is to prefer that format.
A Country station in a market with 70% adults 35-64 scores very differently than a CHR station in the same market. This is where demographic composition meets format preference to produce a realistic share estimate.
Coverage area ethnic composition directly modulates format reach. A Regional Mexican station in a heavily Hispanic market gets an ethnic alignment boost. A Country station in the same market gets an alignment penalty. This modifier was the single largest estimation improvement in v3.0 — it dramatically improved accuracy for stations serving diverse communities.
How many other stations in the same format family serve this market? Rank-based market dynamics distribute audience share realistically: the #1 station in a format gets more than a proportional share, while lower-ranked stations get progressively less. A solo Country station in a small market captures far more of the available Country audience than the fourth Country station in a large market.
Format performance varies meaningfully by US region and season. Country radio over-indexes in the South and Midwest. AC stations spike during the Christmas holiday season. News/Talk surges during election cycles. These adjustments ensure estimates reflect real-world listening patterns rather than flat national averages.
Radio listening peaks during the drive. Census commuting flow data tells us how many workers cross contour boundaries daily — people who live outside the signal area but work inside it (or vice versa). This adjustment accounts for the significant drive-time audience that a purely residential population count would miss.
A complete audience estimate for every station — weekly cume, AQH, and demographic composition — produced entirely from public federal data and established research. No survey required.
Stage 2 tunes and validates the baseline with real listener data. Survey responses and engagement metrics adjust the model — they don't define it. The foundation from Stage 1 means every estimate starts strong, and Stage 2 makes it stronger.
Cardigan sends physical mailers with QR codes to randomly selected households in the market. Recipients register online and complete a friendly, recall-based survey about their radio listening habits. This is not a daily diary — it's a single-session profile of where they can reliably be found on the radio dial.
Survey responses flow directly into the audience model, validating and adjusting Stage 1 estimates across multiple dimensions:
For subscribing stations running Cardigan Active, we incorporate first-party engagement signals that reveal how deeply listeners interact with the brand beyond the broadcast signal.
The TBE score combines the Stage 1 baseline (signal coverage and modeled audience) with digital engagement data and survey responses into a single composite score. This gives advertisers and agencies a holistic view of a station's true reach and engagement — not just how many people can hear it, but how many people actively interact with it.
Every survey response makes every future estimate better. As response data accumulates in a market, the model learns and self-corrects. Markets with more survey data receive higher confidence ratings. The system gets smarter over time — not by changing the methodology, but by feeding it more ground truth.
Traditional ratings (Nielsen/Arbitron) start with a small diary sample — roughly 500 people — and extrapolate everything from that. If the sample is flawed, biased, or unrepresentative, every number built on it is flawed.
Cardigan starts from the opposite direction:
Cardigan is built on authoritative public data supplemented by proprietary research. Every data source is documented and auditable.
Licensed facility parameters, certified service contour coordinates, tower locations, power levels, and antenna patterns.
FederalTract-level population, age, gender, ethnicity, income, and household data from the American Community Survey.
FederalMetropolitan and Micropolitan Statistical Area boundaries from the Office of Management and Budget.
FederalWorker residence-to-workplace flow data for drive-time audience modeling across contour boundaries.
FederalInfinite Dial and Share of Ear studies provide age-specific radio listening rates and media consumption data.
ResearchDecades of industry research on demographic preferences by radio format, updated continuously.
ResearchQR-code mailer surveys with in-market listener responses on station preference, listening habits, and device usage.
ProprietaryFirst-party digital signals from Cardigan Active subscribers: website pixels, contest data, social metrics.
ProprietaryEvery Cardigan estimate includes a confidence rating that tells you how much data supports the number. More data sources = higher confidence.
| Level | What It Means | Data Behind It |
|---|---|---|
| High | Strong directional confidence. Suitable for media planning with appropriate margins. | Stage 1 baseline + survey responses + engagement data from Cardigan Active |
| Good | Reliable estimate with survey validation. Useful for competitive analysis and market positioning. | Stage 1 baseline + survey responses (no engagement data) |
| Moderate | Solid baseline with partial validation. Good for market-level trends and share estimates. | Stage 1 baseline + limited survey or engagement data |
| Baseline | Model-only estimate from public data. Directional intelligence — useful for markets with no survey coverage. | Stage 1 baseline only (contour + census + format affinity + competition) |
Every estimate in Cardigan can be traced back to its component steps. The Prediction Timeline shows you exactly how each number was built — which contour was used, what population was counted, which adjustments were applied, and what the final result means.
Enter any station callsign and see the complete step-by-step walkthrough of how its audience estimate was generated — from signal contour through final adjusted numbers. Every step is labeled as Stage 1 (baseline) or Stage 2 (survey enhancement) so you know exactly what data drives each component.
View Prediction TimelineWe actively encourage feedback on our estimates. If you see data that appears incorrect — a station with the wrong format, a coverage contour that doesn't match reality, or an audience number that seems off — please report it. Every correction makes the entire system more accurate.