NEW YORK--(BUSINESS WIRE)--The Council for Research Excellence (CRE) today reported findings from a new study, "Modeled Consumer Data: Validating the Targets," at the Data & Marketing Association's annual &THEN Conference in New Orleans. The CRE says the study is a step toward bringing transparency to the science of modeled consumer data.
The study utilized data supplied by research partner AIG Travel Insurance, recruited with assistance from the Association of National Advertisers (ANA), to examine how closely the modeled/lookalike targets developed by demand side platforms (DSPs) reflect marketers' true target audiences. The study was conducted for the CRE's Big Data Committee by Pre-Meditated Media.
Presenting the study results at the &THEN event were Gerard Broussard, Principal, Pre-Meditated Media; Newcombe Clark, Global Director, Rapid Learning Lab, AIG; and Stacey Schulman, EVP of Strategy & Analytics for Katz Media Group and Chair of the CRE's Big Data Committee.
"This study brings an important level of transparency to a very opaque process," Schulman said. "Not enough is known about how consumer models are constructed and how accurately they depict intended target consumers. This study is a significant step in pulling back the curtain to enable the industry to better understand and work to improve modeling. On behalf of the CRE, I would like to thank AIG Travel Insurance for partnering with us on this important project."
The CRE chose to explore this issue because the use of modeled consumer data, which combines media audience data with consumer transactions to inform media targeting and measure ad effectiveness, has dramatically accelerated in recent years. Modeling is particularly prevalent in digital marketing, and its use is expected to become more pervasive in television.
To determine how closely modeled targets reflect marketers' true target audiences, AIG participated in an in-market display advertising test of modeled targets executed across six DSPs. Each DSP created modeled target cookies from an anonymized seed file sourced from AIG's customer records. Each DSP served three million unique modeled target ad impressions during a five-week period.
The study sought to answer the following questions:
- To what extent do modeled lookalike targets depict a marketer's true consumer target?
- To what extent does modeled target quality vary across ad tech firms that provide modeled/lookalike targeting services in the digital space?
- Does more accurate modeled targeting translate into superior performance on driving transactions?
- What are the critical questions to ask modeled targeting vendors in order to understand the likelihood that their solutions will reflect desired consumer targets while improving advertising performance?
Key findings and insights included:
- All DSP modeled targeting eclipsed AIG performance benchmarks based on non-programmatic display advertising despite wide swings in match rates. The lookalike-targeted ads surpassed the AIG return on ad spend (ROAS) benchmark by nearly 3x.
- One of the six DSPs excelled, with a match rate nearly double (185 index) the average. But even DSPs with the lowest match rates still achieved ROAS on par with others. Low AIG Travel Insurance match rates paired with strong ROAS suggests these DSPs were credited with significant sales from consumers with no prior AIG purchase history.
- While engaging modeling targeting partners, it is advised to inquire about their experience in the ad category under study, modeling processes, data quality and lookalike model validation procedures. Such discussion will enable the marketer to get a sense of the likelihood that the DSP's modeled target solutions will resemble the marketer's target audience.
"The findings in this study are very valuable to us, and in a broader sense we were very pleased to be able to contribute to industry learnings in what to date has been a somewhat mysterious discipline, making it difficult to verify true performance,” said AIG's Clark. "Without the ability to verify, it’s challenging to trust and invest. The unique way this study was conducted allowed for almost full transparency and measurement of true quality and performance. The results point to the value of collaboration and running concurrent campaign flights with DSPs as each provided strong individual performance and also collectively attributed to overall ROI lift. These processes should become not just best practices, but also global standard operating procedures."
Pre-Meditated Media's Broussard added, "The results of this study are very encouraging, but not enough is known about the how or why. For example, what was special about the one DSP's modeling that enabled it to far outpace the others in terms of ROAS, match rate and consumer target profile? The study illustrates the need for marketers to push for transparency from partners in providing more detail about how they create and validate their consumer models. Additional industry studies could provide more insight and benchmarks on this extremely important topic."
The CRE, funded by Nielsen, operates as an independent organization consisting of senior-level research professionals with the shared goal of advancing the knowledge and practice of methodological research on audience measurement.
For more information, please visit: http://www.researchexcellence.com/