Google Marketing 2020

Creative Audit Methodology

Google's creative audit is an analysis of inclusion dimensions and metrics for marketing images and videos that launched in the U.S. in 2019. Results from the audit are used to understand year-over-year improvement and create annual goals for representation and inclusion in our marketing creative.

Dimensions we looked for in our creative audit

This year, we’re sharing five key insights from the creative audit. However, for greater transparency in the process, here’s a look at all eight dimensions of inclusion that feed into our audit.

Note: These eight dimensions are not intended to be fully inclusive of all possible identities, but they were reviewed internally by people of various identities and experiences. We also acknowledge that many dimensions of identity are not visible to others. A limited set of dimensions was chosen to support measuring representation at scale. Our hope is to understand whether our intent to be inclusive is perceived through external evaluations of our marketing creative.

Age

Perceived presence of people in the age groups of children, teens, young adults, adults, and older adults.

Disability

Perceived presence and representation of people with various types of apparent and non-apparent disabilities.

Gender

Perceived presence and representation of women, men, and gender expansive individuals.

LGBTQ+

Perceived presence and representation of LGBTQ+ people (with the understanding that LGBTQ+ stands for lesbian, gay, bisexual, transgender, and questioning or queer).

Military status

Perceived presence of Veterans or people who are serving in the military.

Race and ethnicity

Perceived presence and representation of Asian, Black, Indigenous, Latino / Latinx, or white people.

Skin tone

Perceived presence of a diverse spectrum of skin tones from dark to light.

Socioeconomic status

Perceived presence of low-income, middle-income, and high-income socioeconomic categories.

Methodology used in our analysis

The Geena Davis Institute on Gender in Media conducted human-evaluated research on 1,400+ images and videos from our top campaigns by budget that launched in the U.S. in 2019. They analyzed the dimensions listed above as well as an expanded list of dimensions.

In addition, evaluations were led by an internal Google team to analyze 6,000+ marketing images and videos that launched in the U.S. in 2019. We worked with over 1,500+ external evaluators to help us understand which identities and intersectionalities are perceived in our marketing. We acknowledge that there is a human margin of error inherent in this analysis, since perception varies based on an individual’s collective identities and experiences, and since many dimensions of identity are not visible or perceivable by others.

Machine learning analysis was led by an internal Google team on 8,000+ videos and on 60+ Google-owned YouTube channels. This analysis focused on data from videos uploaded in 2019 and 2020. To measure representation in video content, the models detect face attributes for use in analysis: perceived gender expression, age, and visual speech detection. You can learn more about the methodology for a similar machine learning analysis on this page.