A single-source panel is a recruited sample of households or individuals whose exposure to advertising and whose subsequent behaviour, usually purchasing, are recorded from the same people over the same period. The design exists to answer a question that separate surveys cannot: did the people who saw a campaign go on to buy more than those who did not? When television ratings come from one sample and shopping data from another, the link between the two has to be estimated. A single-source panel observes it directly. The term is also used, increasingly, for panels that measure one set of people across several media at once, such as linear television and online video, so that duplicated reach can be counted rather than modelled.

How a single-source panel works

The mechanics have four parts. A research company first recruits a sample designed to mirror a population, typically by region, household size, age and income, and weights the results to official population estimates. It then fits each home with exposure measurement: a people meter on the television, on which each viewer registers presence, software or router meters for phones, tablets and computers, and sometimes a portable or wearable meter that detects audio codes embedded in broadcasts. Third, it records outcome data from the same homes, historically through barcode scanning of purchases at home, ID cards presented at checkouts, or diaries. Finally, both streams are joined at the household or person level by the panel's own identifier, not by a probabilistic match.

A worked example shows what this buys. Suppose a cereal brand airs a four-week television flight. In a single-source panel of 2,000 homes, the analyst can split households into those that were exposed to at least one spot in the seven days before a shopping trip and those that were not, and compare the brand's share of category purchases between the two groups. John Philip Jones formalised this comparison in 1995 as STAS (short-term advertising strength): the brand's purchase share among households that saw an ad in the preceding week, indexed against its share among households that did not. An index of 120 means exposed households bought the brand at a rate 20 per cent higher.

The cross-media variant replaces purchases with a second media stream. The Origin panel built by Kantar for ISBA, the UK advertisers' association, is described in ISBA's August 2024 FAQ as "a single-source measurement panel consisting of a group of households or individuals who have agreed to share personal details". Its 2,500 homes carry people meters for linear television, mapped to commercial spot times in the manner used by Barb, the UK's television currency, alongside measurement of online video and display. In such systems the panel rarely produces the published numbers on its own. It acts as a calibration panel, or "truth set", used to train a virtual ID (VID) model that labels census-level impression logs from publishers with estimated people, from which deduplicated reach and frequency are calculated.

Panel sizes are modest, typically between 1,000 and 5,000 homes, as the history below shows.

Origin and evolution

The idea is usually credited to Colin McDonald of the British Market Research Bureau (BMRB), who in 1966 ran what is often described as the first quasi-single-source study, asking the same respondents to keep both viewing and purchase diaries across nine product fields, including washing powders, cereals, tea and toothpaste. Primary records of the exact sample are not public.

Electronic versions arrived in the United States. AdTel was an earlier attempt, but IRI's BehaviorScan, launched in January 1980 in Marion, Indiana, and Pittsfield, Massachusetts, became the reference design. IRI supplied scanners to 15 supermarkets in the two towns, recruited about 2,000 households in each, and attached devices to their televisions that allowed it to substitute one commercial for another in cable broadcasts. Sources disagree on IRI's founding year: Wikipedia gives 1979, while company histories give 1977.

Arbitron followed with ScanAmerica, tested in Denver from late 1986 and rolled out nationally on November 4, 1991, after CBS signed on. About 1,000 households scanned purchases with a handheld wand while a people meter recorded viewing. Arbitron committed $125 million to the service and withdrew the national product in 1992, after about ten months, citing "a lack of a viable subscriber base". In 1995 Jones published "Single-Source Research Begins to Fulfill Its Promise" in the Journal of Advertising Research, reporting strong short-term effects for about 20 per cent of brands studied and modest effects for another 30 per cent.

The most ambitious attempt was Project Apollo, developed by Arbitron and Nielsen from 2005. It combined Nielsen's Homescan purchase panel with Arbitron's Portable People Meter, and seven advertisers including Procter & Gamble, Unilever, Kraft and Walmart sat on its steering committee. The companies ended it in February 2008, stating that they "did not secure sufficient client commitments".

After Apollo, the market moved toward matched rather than true single-source data. Nielsen and Catalina Marketing formed a 50-50 venture on December 14, 2009, to link Nielsen's television and internet panels with purchase data from more than 50 million shoppers in Catalina's loyalty network. TiVo agreed to buy TRA in July 2012 for about $20 million; TRA matched set-top box data from 1.5 million homes with supermarket loyalty-card records. In Europe, GfK announced a UK Media Efficiency Panel of 10,000 households on November 30, 2009, with purchases scanned at home and web exposure tracked by software, part-funded by Google for pilot studies.

Why it matters for marketers

Single-source panels sit underneath two numbers that drive budgets: sales impact and deduplicated reach. Purchase-linked measurement now usually runs on loyalty-card or receipt data at far larger scale, as described in PPC Land's explainer on sales lift, but the panel logic of comparing exposed and unexposed buyers from one source remains the template. Ownership of those assets keeps changing. Circana agreed in August 2024 to acquire NCSolutions, the Nielsen and Catalina venture, and Infillion bought Catalina in February 2026 with data covering 130 million US households.

On the media side, advertiser-led cross-media programmes treat a single-source panel as the independent referee between platforms that report their own audiences. The WFA's Halo framework, which combines single-source panel and digital census data, underpins both Origin in the UK and Aquila in the US.

Limitations and disputes

Cost and sample size have defeated most single-source services, according to accounts of ScanAmerica and Apollo. Metered homes are expensive to recruit and retain, and scanning every purchase or registering every viewing session invites fatigue, under-reporting and attrition. A 2,000-home panel splits quickly into cells too small to read for a niche brand or a short flight.

Representativeness is a second dispute. Loyalty-card data covers only participating retailers, and set-top box data skews toward pay-television homes, problems PPC Land reported in the VAB's August 2026 guide. The same guide warns that any bias in a calibration panel carries into the larger dataset it is meant to correct, which matters because modern single-source panels mostly calibrate rather than measure.

Panel size also feeds commercial conflict. ITV's commercial managing director, Kelly Williams, objected that Origin proposed "a much smaller panel" of about 2,000 households than Barb's, warning of "two different currencies in the market", according to The Media Leader. ISBA replied that it was using linear data from Kantar's 2,500-home panel. Finally, matched datasets marketed as "single source" are not panels in the strict sense: they join records from different collection systems, so match rates and identity errors replace sampling error as the main uncertainty.

Single-source panel versus adjacent terms

Data fusion. Fusion statistically joins two separate samples, such as a TV panel and a purchase survey, by matching respondents on shared characteristics. A single-source panel avoids that step by measuring both behaviours in the same people.

Census data. PPC Land's census data explainer describes a record of every event in a universe, such as all impressions served. Census data has scale but no demographics; the panel supplies the people.

Syndicated research. Syndicated research is a business model, a study sold to many subscribers. Some single-source panels are syndicated, others built for one client.

Single source of truth. A data-governance phrase for one authoritative record of each data element. It has no connection to panel research.

Recent developments

The cross-media panels moved from trials to products in 2026. Aquila, whose Kantar-built panel PPC Land described as a single-source calibration panel when it added Samba TV streaming data in October 2025, began a first phase covering Meta, YouTube and linear television within 30 days of March 26, 2026, with 27 advertisers signed, according to MediaPost. TikTok, Amazon and iHeartRadio are scheduled for the second half of 2026.

The panel supplier itself changed hands. Kantar Media relaunched as Fifty5Blue on February 25, 2026, after H.I.G. Capital completed its acquisition in August 2025, summarising its model as "panels for rigour, big data for scale". On August 3, 2026, ISBA and Fifty5Blue announced exclusive talks for a minority stake in Origin, with an agreement targeted for September, according to Daily Research News Online. No signed agreement had been reported in the sources reviewed as of October 18, 2026.

Panels' dependence on platform cooperation was exposed in January 2026, when YouTube forced Barb to halt its UK channel measurement after a legal threat over audio-matching technology used in panel homes. In the US, Nielsen's Big Data + Panel system combines a 42,000-home panel with device data from 45 million households, the same division of labour.

Timeline

  • 1966 - Colin McDonald at BMRB runs a diary study recording viewing and purchasing for the same respondents.
  • January 1980 - IRI launches BehaviorScan in Marion, Indiana, and Pittsfield, Massachusetts, with about 2,000 households per town.
  • Late 1986 - Arbitron begins testing ScanAmerica in Denver.
  • November 4, 1991 - ScanAmerica rolls out nationally after CBS signs on.
  • 1992 - Arbitron drops the national ScanAmerica service after about ten months.
  • April 1995 - John Philip Jones publishes single-source STAS findings in the Journal of Advertising Research.
  • 2005 - Arbitron and Nielsen begin developing Project Apollo.
  • January 2006 - Apollo's national pilot reaches more than 11,000 people in 5,000 households.
  • February 2008 - Project Apollo is shut down.
  • November 30, 2009 - GfK announces a 10,000-household UK Media Efficiency Panel.
  • December 14, 2009 - Nielsen and Catalina Marketing form a 50-50 joint venture.
  • July 2012 - TiVo agrees to buy TRA for about $20 million.
  • September 22, 2021 - Kantar is named panel supplier for ISBA's Origin.
  • June 2024 - The ANA announces Aquila.
  • August 2024 - ISBA's FAQ describes Origin's 2,500-home single-source panel; Circana agrees to buy NCSolutions.
  • March 3, 2025 - Aquila's calibration panel is installed in more than 1,000 homes.
  • July 9, 2025 - Origin moves to expanded availability.
  • August 2025 - H.I.G. Capital completes its acquisition of Kantar Media.
  • January 28, 2026 - Barb confirms a pause in YouTube channel measurement.
  • February 25, 2026 - Kantar Media relaunches as Fifty5Blue.
  • March 26, 2026 - The ANA says Aquila's first phase will roll out within 30 days.
  • August 3, 2026 - ISBA and Fifty5Blue announce talks on a minority stake in Origin.

Summary

Who. Research companies build and run single-source panels: BMRB, IRI, Arbitron, Nielsen and GfK historically, and Kantar Media, now Fifty5Blue, for Origin and Aquila. Advertisers, their trade bodies ISBA and the ANA, and consumer goods marketers are the main buyers.

What. A recruited sample of households or people whose ad exposure and purchases, or exposure across several media, are measured together from the same individuals, removing the need to fuse separate datasets.

When. The concept dates to Colin McDonald's 1966 diary study. Electronic versions began with BehaviorScan in January 1980. ScanAmerica closed in 1992 and Project Apollo in February 2008. Cross-media calibration panels for Origin and Aquila followed from 2021, with Aquila's first phase rolling out from spring 2026.

Where. The model originated in the UK and was industrialised in the US. Current flagship deployments are Origin in the UK, with 2,500 homes, and Aquila in the US, which passed 1,000 homes in March 2025.

Why. Separate samples cannot show whether the people who saw an ad behaved differently. A single-source panel observes that link directly, which is why advertisers fund it as an independent check on platform-reported audiences, despite its cost and small samples.