How Effective is Black-Box Digital Consumer Profiling and Audience Delivery?: Evidence from Field Studies

Privacy and Security and Innovation and Economic Growth

Article Snapshot


Nico Neumann, Catherine Tucker and Timothy Whitfield


Forthcoming in Marketing Science - Frontiers


Little is known about how data brokers create consumer profiles for use in marketing. Analysis shows that digital profiles purporting to identify consumers by age and gender are often inaccurate. Advertisers globally waste $7 billion annually buying such profiles.

Policy Relevance

When purchasing profiles, firms should ask for evidence of the validity of the data broker’s methods. Programs to standardize and verify broker methods would improve accuracy.

Main Points

  • A Demand Side Platform (DSP) sells advertising services to advertisers, who provide the DSP with a description of the desired target audience for the ad.
  • DSPs need data on each user to target the ad, obtaining this data from data brokers; data brokers consolidate their data in a database maintained by a data management platform (DMP).
    • One DMP has links to many data brokers.
    • One DMP is usually integrated with multiple DSPs.
  • One study of an ad campaign targeting males aged 25-54 found that the average accuracy of the data brokers was 59%; using the data brokers resulted in targeting that was 184% more accurate than a randomized campaign, but targeting cost 238% more.
  • In a second study, three data brokers were asked to identify the age and gender of visitors to a given site; the brokers’ accuracy in identifying gender ranged from 55% to 85%, and their accuracy in identifying age ranged from 6% to 46%.
  • A third study of 14 data brokers showed that the accuracy of gender classification varied from 25.7% to 62.7%, while the accuracy of age classification varied from 17.8% to 36.7%.
    • Classification by age is more efficient than classification by gender, which is surprising, as age classification has a greater chance of being wrong.
    • The average accuracy of data brokers was only 35.8%.
  • Several factors affect data brokers’ accuracy.
    • Accuracy varies by country, with users in the United States being identified correctly more often than users in the United Kingdom or Australia.
    • The presence of children reduces accuracy, as it is more likely that multiple users are sharing a device.
  • The inaccuracy of profiling results in $7 billion of wasted expenditures every year; also, when targeting is less than 65% accurate, consumers are more likely to install ad blockers.
    • Businesses should rely on their own data, or ask data brokers to verify the accuracy of their methods.
    • Industry standards and certification processes for data brokers would improve accuracy.

Get The Article

Find the full article online

Search for Full Article