Nielsen Optimizes Cross-Platform Reach Measurement

Since 2019 the World Federation of Advertisers the Association of National Advertisers the Incorporated Society of British Advertisers and the advertiser associations in 58 other countries have been focused on being able to accurately and precisely measure the reach and frequency of their ad campaigns across television and digitalNielsen ONE Ads was launched on January 11 2023 to meet this need in the US Tens of thousands of campaigns have been measured so far Because my consulting for Nielsen grants me access to their data in this column Iamp8217ve examined what the industry can learn from the patterns reported here from across the first 45000 campaigns measured Weamp8217ve noted that dispersion across more publishers and platforms enables a campaign to gain as many as 20 more reach points compared to campaigns which are very focused on just a few of the most obvious media choices Weamp8217ve shown that linear is still the essential reach driver but CTV mobile and computer need to receive significant allocations as well In our reporting here made possible through my consultancy for Nielsen so far weamp8217ve concentrated on reporting findings rather than writing about methodologyIn this article Iamp8217m going to focus on some methodology work that Nielsen has done which I feel is worthy of public attention As a researcher myself I cannot help but call out what I consider to be great accomplishments in data science Nielsen presented this methodology at the ARF annual AUDIENCExSCIENCE 2025 conference in March and has just written it up as a paper soon to be released I will draw excerpts from that paper as a preview here for a general audience of media leadersFirst letamp8217s note a few special problems in measuring reachfrequency across an immensely fragmented sea of diverse media types in a way that addresses the business and workflow needs of all players in the ecosystem advertisers agencies and media sellersFor any audience research company in any country there will be a currency that until now has been for linear television Linear is now only one of a number of types of television VOD streamingCTV Addressable TV TV on mobile devices TV on computers IPTVATSC 30 and a zoo of digital types The methods of measuring crossplatform must cover all of these That methodology cannot be identical to currency And yet if its results do not correspond to currency for linear and streamingCTV the marketplace breaksOutcomes are where the market is moving In order to ensure the best ROAS and branding outcomes the new paradigm requires inflight reoptimization meaning dropping certain nonperforming ads and media and shifting budgets to what is working Thus a crossplatform measurement system must be able to work from initial data collections which are likely to be sketchy Remember that the ratings which used to average 180 in primetime are now in the vast majority of cases down in the 0XXX three decimal places range No matter how big the panel and the big data array with fractionated audiences the first day of a campaign is going to provide data that is going to misrepresent the cumulative findings across the first week much like the early results of an election do not accurately predict the final tallyThese are among the realworld concerns which the Nielsen data science engineering and product teams took into careful consideration in coming up with the elegant methodological solution weamp8217re about to describe Iamp8217ll now be providing excerpts from the paperA The Foundation Nielsen39s SingleSource Panel The Ground TruthAt the heart of Nielsen ONE Ads39 accuracy is Nielsen39s robust singlesource panel This isn39t a theoretical model it39s a direct observation engineRich Comprehensive Data In the US nearly 27000 TV homes encompassing approximately 67000 panelists are equipped with Nielsen Streaming Meters that capture digital behavior alongside traditional TV meters This provides a holistic view of media consumption at the household and individual levelPersonLevel Insights Crucially mobile and computer devices within these panel homes are mapped to individual panelists This personcentric approach allows for proper attribution of digital impressions accounting for nuances like device sharing or coviewing within a householdDirect Observation of Duplication By combining metered TV data with matched digital impressions either through streaming meter data or panelist matches for certain walled gardens the panel directly captures respondentlevel ad exposures across platforms This provides an unbiased empirical source of truth for how actual audience duplication occurs for individual campaigns with data processed dailyThis singlesource panel is the bedrock providing the direct observational data essential for truly understanding crossplatform behaviorB Dynamic Deduplication Blending Direct Observation with Modeled Intelligence via Bayesian InferenceWhile direct observation is vital Nielsen ONE Ads enhances this with intelligent modeling leveraging the power of Bayesian inference to combine multiple streams of information for the most robust estimateTwo Key InputsDirect Observations The quotPanelquot Step quotLikelihoodquot Campaignspecific ad exposures captured directly from the singlesource panel provide granularity and reflect the unique characteristics of that campaignModeled Expectations The quotExpectationsquot Step quotPriorquot Nielsen leverages rich historical data from tens of thousands of crossplatform campaigns run since early 2022 This data is used to train models that provide expected duplication rates offering stability to the estimates especially when panel observations for a specific campaign might initially be sparseFigure 3 Duplication rates depend on both modeled data and direct observations weighted based on the amount of signal in the singlesource panel These are recalculated daily for each separate campaign throughout its reporting lifeThe Bayesian Advantage Instead of relying on just one source Bayesian inference intelligently weighs and combines both the campaignspecific panel data likelihood and the historical modeled data prior The result is a quotposteriorquot distribution amp8211 a final more accurate and stable estimate of duplicationFigure 4 Example likelihood prior and posterior distributions for a given crossplatform duplication rate The use of a Bayesian framework ensures that both the historical data and campaign panel data are used to their fullest in the estimation of duplication ratesDynamic Weighting This isn39t a static calculation The influence of the panel data versus the historical model is dynamicFor campaigns with a large volume of panel observations high panel signal the system leans more heavily on the direct campaignspecific dataFor smaller campaigns or at the beginning of a campaign when panel sample observations are still accumulating the historical models provide a stable and reliable starting pointThis weighting continuously updates daily as more panel data becomes available for a campaignThis dynamic datadriven approach ensures that Nielsen ONE Ads maximizes the utility of all available information balancing the stability of historical trends with the specific nuances of each individual campaignC Ensuring Coherent and Consistent Reporting The Alignment StepObtaining accurate duplication insights is one part of the equation ensuring these insights are reported consistently and make logical sense alongside established platform metrics is equally critical Nielsen ONE Ads incorporates a rigorous alignment processAlignment with Platform Totals The deduplicated audience estimates are carefully aligned with Nielsen39s trusted Television Audience Measurement TAM and Digital Ad Ratings DAR platform totals This ensures that the crossplatform view is coherent with the currency numbers clients already useStatistical Optimization for Consistency To maintain consistency across all reporting levels eg from individual placements up to the total campaign and across different audience segments Nielsen ONE Ads employs sophisticated statistical algorithmsMaximum Entropy This principle is applied at each level of reporting granularity eg quothorizontallyquot across the different screens like TV CTV Computer Mobile It ensures that the estimated duplication rates align with the total audience recorded for each screen in a way that introduces the least possible bias when information doesn39t uniquely determine a distributionLinear Optimization A separate optimization step is then applied quotverticallyquot across reporting granularities This ensures for example that the sum of audiences for individual placements logically rolls up to the site level which in turn rolls up to the total platform level and that these all relate coherently to the total crossplatform audienceThese meticulous alignment and optimization steps ensure that the millions of estimates produced for any given campaign are not only accurate in isolation but are also selfconsistent and interrelated providing a trustworthy and logical foundation for decisionmakingLet39s examine the evidence from actual campaignsA Discrepancies in Unique Audience and Overlap PatternsVisualizing the audience distribution across platforms immediately reveals stark contrasts Consider two masked fourscreen campaigns from 2024Campaign 1 Higher Observed Duplication by Nielsen ONE Ads In this first example Nielsen ONE Ads by observing actual audience behavior identified higher rates of duplication across platforms than the independence assumption would predict This resulted in a total unique audience UA that was 6 smaller than what the independence assumption suggested For a campaign with approximately 134 million total unique viewers as measured by Nielsen ONE Ads this 6 difference translates to about 8 million fewer unique individuals compared to the independence estimate equivalent to the population of New York City or Washington State In this scenario relying on the independence assumption would have significantly inflated the perceived reach of the campaignFigure 6 Sample Campaign 1 Total Unique Audience Comparison between N1 Ads and Independenceamp160 This highlights an example where N1 Ads measures more duplication across platforms than Independence would have naively calculated resulting in a smaller total unique audienceamp160Campaign 2 Lower Observed Duplication by Nielsen ONE Ads In a second campaign from the same timeframe the situation was reversed Nielsen ONE Ads observed lower rates of actual duplication leading to a total unique audience that was 3 larger than the independence assumption For this campaign which had a total unique audience of 217 million as per Nielsen ONE Ads this meant Nielsen reported approximately 6 million more unique individuals than independence would have Here the independence assumption would have led to an underestimation of the campaign39s true reach potentially causing advertisers to undervalue its impact or miss opportunitiesFigure 7 Sample Campaign 2 Total Unique Audience Comparison between N1 Ads and Independence This highlights an example where N1 Ads measures less duplication across platforms than Independence would have naively calculated leading to a larger total unique audienceamp160These individual campaign snapshots illustrate a critical point the independence assumption is not consistently biased in one direction It can lead to either overstatement or understatement of reach making it an unreliable tool for decisionmaking When attempting to understand the duplication between specific components the relative error range is often far greater than at the campaign levelThe new media ecological landscape is far more difficult to precisely measure than it was when some of us first got into the business and it will continue to become more complex from now on with Virtual Reality VR Augmented Reality AR AI Agents that read a personamp8217s EEG and eventually fMRI as widespread as todayamp8217s Apple Watches and Fitbits and bring new media content to that person to maximize desired moods and eventually 3D haptic VR that is a deepfake of reality Keeping up with all of it will be dedicated media researchers who truly care about accuracy and are kept awake at night figuring out solutionsPosted at MediaVillage through the Thought Leadership selfpublishing platformClick the social buttons to share this story with colleagues and friendsThe opinions expressed here are the author39s views and do not necessarily represent the views of MediaVillageorgMyersBizNet

Click to Make Your 2025 Contribution to Support the MediaVillage Education Foundation

Nielsen Optimizes Cross-Platform Reach Measurement

Author

Publish date

Channel

Archived in

Bill Harvey

The Ad Council, Huntsman Mental Health Foundation and Amazon Announce “The Mind Set” to Support Mental Health - Ad Council InSites

More Videos:

Click to Make Your 2025 Contribution to Support the MediaVillage Education Foundation

Nielsen Optimizes Cross-Platform Reach Measurement

Enjoying This Commentary? There's More to Love

Bill Harvey

TAGS:

The Ad Council, Huntsman Mental Health Foundation and Amazon Announce “The Mind Set” to Support Mental Health - Ad Council InSites

The payment was successful

Cookies