Page 21 - Data Transparency White Paper_FINAL_Neat
P. 21
Addressing Biases in Multicultural & Inclusive Identity Data
21
APPENDIX:
QUESTIONS TO ASK THIRD-PARTY DATA PROVIDERS
CATEGORY QUESTION WHY IS IT IMPORTANT?
• It is important to know how the
underlying raw data was obtained. U.S.
What are the sources of data Census records, probability panels/
the provider uses for race, surveys, public records, transactions,
Data Sources searches, social activity, physical
ethnicity, language, gender location visits, cookie data, mobile
identity, age, etc.? event data, proprietary algorithms
– any or all can be the source of a
particular segment.
Does the provider collect
data themselves or • When third parties aggregate data
aggregate it from other from multiple sources, there is often
sources? Has it been sent limited insight and input into the
through a chain of different quality or accuracy of the data.
Data Sources
providers? • If the provider licenses data from
another provider, then they are likely to
Does the provider build its have limited insight into how the data
own identity graph, or simply comes together.
license other providers’ data?
• It is unrealistic to expect providers to
have a 100 percent coverage rate of
the population or specific segments.
What is your audience? What Some have higher coverage of
Coverage is the provider’s coverage for homeowners, credit card holders,
that audience? or smaller-sized households. This
introduces bias when targeting
multicultural consumers due to low
coverage in the underlying data.
• There are different ways providers
define race/ethnicity or language such
as first name, surname, country of
How does the provider define origin, English proficiency, U.S. Census
race/ethnicity, language definitions, etc. There is no standard
Definition (Census-based) definitions for sexual
dominance, and gender or gender identity, acculturation level,
identity? Spanish language dominance, etc.
So, it is important to understand how
these segments are defined by the
provider.
