In general, these errors can be tolerated when they are random in nature. But when they are systematic, consequences for decisions based upon such faulty data can be significant.
Take for example if systematic error sneaks into the samples used for television audience viewing panels. If people who are heavily engaged with television are more likely to consent to participate in a panel measuring viewership, then the ratings will overstate the size of the audience and advertisers will overpay.
But systematic error is a much more significant problem when it comes to representation of minority audiences in research.
A study for the Alliance for Inclusive and Multicultural Marketing (AIMM) recently found that while Whites were adequately and accurately identified 68% of the time in large digital datasets used for target marketing, that “visibility” dropped to 49% for Hispanics, to 28% for African Americans, and to 24% for Asian Americans
The accuracy problem is growing more acute because of fundamental changes in society. The rise of blended families and multicultural households makes their characterization more challenging. The increasing diversity of neighborhoods makes it harder to assign demo attributes based upon geography. Geo-targeting is further undermined by the increasing tendency of younger consumers to move around.
Additionally, ad blockers and more stringent privacy requirements make less reliable the identifiers used to ascribe characteristics to records in many commercial databases.
To address this issue, AIMM recently outlined a series of recommendations. While these do not address the root cause of underrepresentation of racial minorities in audience segments, they do offer up ways to at least begin to isolate the problem:
Push for greater transparency from third-party data providers. Marketers and agencies must insist on seeing the composition of the segments they use and the date ranges when such data were compiled.
Get more serious about data quality. There is a tendency to save on cost by using lesser quality data purely for directional analysis. This might suffice for looking at behavioral segments such as category buyers, but it is inadequate when making investments in studies that serve as media currencies or that lead to consequential business decisions.
Benchmark against self-reported data to ensure better ID assignment validation. While self-reported identity and language preference data can be unreliable, for marketing purposes, they should not be dismissed. Regardless of what a consumer’s birth certificate or DNA may say, it’s how they self-identify that makes products and messages more or less relevant and appealing. Visual verification, as provided by in-person interviews, can add another layer of certainty.
As the industry becomes more data-driven, investments in quality and representation of diversity will become more critical.
The consequences of underinvestment are significant. For example, lack of representation could negatively impact ratings for Spanish language programming, leading it to be undercounted, resulting in reduced rates and advertising.
Ultimately, the undercounting of racial minorities leads to marketers not investing in them. It robs these groups of economic power and the power to guide industry to produce products and programming reflective of their needs.
The marketing industry has made significant commitments to increasing investment for minority-owned businesses and media companies. Those efforts are critical and are to be applauded.
But failure to increase investment in more diverse representation in market research will reduce the impact and effectiveness of all other diversity initiatives.