big data

Forecasting & Optimizing Reach in a PII Compliant Measurement Ecosystem

Spencer LambertVP, Product & Partnership Success, datafuelX

Matthew WeinmanSr. Director, Advanced Advertising Product Management, TelevisaUnivision

Reach and frequency planning requires access to unique viewership data, which has become increasingly restricted due to identity restrictions. However, challenges exist with panel-only measurement, including the undercounting of Hispanic and Spanish language coverage, stated Matthew Weinman (TelevisaUnivision). Panel data undercounts Hispanics audiences by upwards of 20%, even for broad demographics. The benefits of big data exist across audience planning, viewership measurement and outcomes. Excessive frequency can be limited while maintaining or expanding reach, as well as improving ROAS. However, there are barriers to working with big data, including PII compliance. Additionally, the size and scale of big data leads to lengthy ID forecast times and computing costs. Spencer Lambert (datafuelX) presented details of their approach to ID-level forecasting which included their reach and frequency clustering methodology. Key takeaways:
  • Advantages of clustering methodology over identity methodology for reach and frequency:
  • Efficiency and accuracy: Delivers comparable accuracy metrics
  • Lower error rates: Seven percent for cluster reach forecasts vs. 20% error rate on identity-scaled reach forecasts
  • Cross-platform reach and frequency: By scaling cluster assignments to digital IDs, this methodology can empower cross-platform management and optimization
  • Lower compute time and costs
  • PII compliant: Preserves the use of identity-level planning

Download Presentation

Member Only Access

Optimizing Big Data + Panel Measurement Through Calibration

David KurzynskiSVP, Data Science, Nielsen

Kyle PoppieVP, Data Science, Nielsen

It is challenging to measure the smaller audiences of local TV and measurement challenges include false zero audience metrics and instability. Kyle Poppie (Nielsen) reviewed the evolution of local TV measurement, and this presentation demonstrated how Nielsen’s approach enables accurate measurement. Calibrating big data to a probabilistic panel controls for biases in the big data population that cannot be accounted for by weighting alone. The panel provides accurate and unbiased measurement at aggregate levels while big data provides greater coverage of granular behavior. An example demonstrated how the calibration of panel data and big data resulted in a more accurate weighted audience size. David Kurzynski (Nielsen) presented a case study that applied calibration to live data from a secondary station in New York. The improved result included fewer zero ratings and smoother trends. Key takeaways:
  • The goal of calibration is to achieve local TV measurement that provides accuracy and stability for audience levels and audience flows.
  • Both big data and panel data are critical as inputs to calibration to achieve these goals. Audience levels are informed by both big data and panel homes, and audience flows are influenced by big data.
  • Relative and total errors decrease as a result of calibration compared to panel-only currency.

Download Presentation

Member Only Access

How AI Will Make Us Smarter

During a recent ARF Town Hall, Professor Russel Neuman explained why he thinks AI will have primarily positive effects. It will make us smarter and more productive. Furthermore, he does not think new regulations are needed to protect humanity from negative consequences of AI development. Read more »

Next Generation Artificial Intelligence


Professor Russ Newman of New York University does not believe that AI will cause humanity’s extinction. Instead, it should help enhance human intelligence and productivity and our quality of life. After putting the AI revolution into historical context, Prof. Newman discussed aligning AI with human values. At our current stage, he believes the regulatory mechanisms in place are sufficient. He explained how large language models work, what allowed them to come into existence and their future impact, describing the effect on marketing and advertising, as well as what the individual user experience will be like. A democratizing, hyper-personalized experience could take place where AI agents advocate on their owner’s behalf and negotiate each transaction with their owner’s preferences in mind. Over time, he sees a great diversification of models coming into being. Historically speaking, each groundbreaking technology that changed the world has been a net gain for humanity. What makes AI different is that if applied correctly, it could make us smarter. The question is, if AI gives us exceptional advice, will we take it?

Member Only Access

Advertising Challenges

Editor-in-Chief Colin Campbell’s editorial, in JAR’s December 2023 issue, outlines challenges faced by advertising practitioners and researchers.

Read more »

Making the Right Impression


Key Takeaways

Paul Donato – Chief Research Officer, ARF Pete Doe – Chief Research Officer, Nielsen Pete Doe brought some clarity to how Nielsen currently approaches linear TV measurement and how it will evolve throughout 2024 in this detailed presentation, describing Nielsen’s integration of big data with panel data in its national TV measurement, participation in auditing and accreditation, exploration in defining impressions and conversations with the industry about time requirements and duration weighting. Other topics discussed in the Q&A that followed covered definitions of calibration and campaign reach measurement, panel adjustments for STB/ACR data, personalization and the differentiation between 30-second ads, 15-second ads and 60-second ads in terms of equivalization and measured impressions. These are selected excerpts from the session’s presentation and Q&A:
  • While big datasets are necessary to capture the fragmentation in the market, panel measurement—with its details on the persons viewing and devices being used—is essential to create a holistic view of audiences. Nielsen’s philosophy does not prioritize one over the other; instead, each informs the other.
  • After listening to industry publishers, agencies and clients, Pete assured the audience that Nielsen will still be offering C3 and C7 metrics in addition to new offerings of individual commercial metrics as of September 24th, 2024. He outlined a three-step process in Nielsen’s overall approach to its big data solution, starting with providing one year of impact national STB (set-top box) data that will then be audited by the MRC and submitted for accreditation. Pete noted that some clients were open to using non-accredited data in the interim, with buyers and sellers agreeing to available data that enables transactions.
  • Nielsen’s currency roadmap for 2024 begins with the currently available data streams that include both panel-only C3 and big data. They are planning to extend their national big data to include Comcast’s STB data calculated from sub-minute crediting in January and fully release their new currency combination of panel and big data, produced to C3 and C7 standards, in September, subject to auditing and accreditation processes. Pete also provided details on Nielsen’s approach to local TV measurement by introducing a calibration methodology, along with top line national demo findings in age groups and increases in Hispanic and Black audiences from Q1 2023.
  • Pete addressed the importance of having a consistent definition of an impression and how Nielsen worked to achieve more granularity in measurement with the sub-minute level of data. Referencing the MRC’s cross-media measurement standard and the continued debate around time requirements (at least two consecutive seconds) and duration weighting, he said Nielsen found no complete consensus from different sides of the industry, although there seems to be more support for two continuous seconds without duration weighting. Nielsen’s exploration in defining impressions assumed that equivalization as a kind of duration weighting will be assessed as deals are made.
  • Nielsen compared the impact of 1s, 2s and 5s using their sub-minute panel plus big data measurement against panel data and the average commercial minute, and, when adding duration weighting, found significant differences in impressions across varying age groups, households and day parts.
  • In terms of deals, one of the benefits of moving from the average commercial minute in a program to individual commercial metrics is the ability to look at the position in the commercial pod. In an example from a daytime broadcast show, Pete illustrated how first-in-pod ads typically deliver a higher audience than the rest of the ads in the pod, finding 99 percent of ads in the first pod indexed higher with 18 percent higher impressions than the average across 160 placements.
  • Nielsen’s national measurement’s “big data” encompass 30-35 million homes including Comcast, DirecTV and DISH return-part-data (RPD) from STBs. Smart TV ACR data from Roku and Vizio adds to the 30-35 million total with some overlap. In local markets, Nielsen does not currently use smart TV data as local stations are not all measured or supplied in its numbers so they focus instead on RPD augmented with Charter data. Because of its deals with DirecTV and DISH, Nielsen has a presence in every market.
  • Nielsen has streaming meters in about 50% of the homes in its national panel currently and is focusing on building those numbers. It also has local CTV measurement capability.
  Nielsen’s key takeaways:
  • Panel+big data means higher audiences, better stability, fewer zero ratings.
  • Overall patterns of viewing are pretty consistent between panel and panel+big data.
  • Two-second qualifier increases available impressions, while duration weighting deflates them.
  • Individual commercial minute data enables pod position considerations in deals.

Download Presentation

Watch Presentation

Member Only Access

THE LAST WORD: Wednesday

There was a combination of both upbeat and downbeat takeaways from the anchor commentators on the last day of the conference.