35.5 Frontier Papers
35.5.1 (Neumann, Tucker, and Whitfield 2019)
19 data brokers , 6 buying platforms, 90 third-party segments
Descriptive Analysis
Study 1:
Examine performance of an ad campaign with the support of data (to target customers)
Automated system can only delivery 59% to the target market.
Audience accuracy varies between platforms.
Study 2:
Examine the optimization of DSPs (Demand-side platforms) for selecting data sources and ad placements.
Delivering performance = f(audience selection, quality of the profiles by data brokers, and other factors).
This study only focuses on the quality of profiles by data brokers.
Optimization is worse than random selection (because average accuracy of identifying the true subject is 24.4% which is less than 26.5% according to the natural distribution of the two attributes - age and gender).
Households with children significantly reduce the performance accuracy (due to potential usage by multiple members)
Study 3:
Audience interest-based data are the new type of target (besides age and gender)
Sports interested
fitness interested
travel interested
High accuracy for this interest-based (but still variation by data brokers)
Cost-benefit analysis
Cost = fixed (third-party audience info) + variable costs (cost-per-mille of online ads)
Ad optimization is more costly than banner (about 151% more), but compared to the gain, third party solution is still economical.