Disclosure Best Practices vs. what Amazon and Google Do

A few weeks ago Google (Alphabet) published their 10th annual diversity report. In many ways, their trend toward transparency and usability should be lauded:

  • Disclosure of self-id data—LGBTQ+, disability, and veteran representation was added in 2019
  • The share of attritions, reported by region and broken down intersectionally by gender and race/ethnicity, was added in 2022 (previously it was reported as an index)
  • They have publicly disclosed their eeo-1 data since 2014
  • They have reported race/ethnicity data outside the US for the last 4 years
  • They make all the underlying raw data going back to 2014 available for analysis in BigQuery

However, in their 2019 report, they started using what they call the ‘plus system’ for race/ethnicity categorization “because multiracial people are ‘plussed in’ to each racial category they identify with.” That means that when somebody identifies as two or more races/ethnicities, they are counted in both categories.

For example, for their overall workforce in 2023 if you add up all the race/ethnicity categories you get 105.1%. In fact, none of their race/ethnicity data adds up to 100% because everyone who identifies with more than one race is double-counted, and this makes it nearly impossible to compare their data to other companies because nobody else uses this system….except Amazon.

Amazon has similarly been publishing workforce representation data for 10 years, and not to be outdone, began using this ‘plus system’ in last year’s report. Similarly adding up all races/ethnicity for their overall workforce you would get 104.7%.

I get the premise – increased transparency and granularity of how people self-identify instead of fitting everything into neat little boxes, but doing this instead as opposed to a supplementary disclosure makes it very difficult for the readers of these reports to actually use them.

Citi is a great example – in each of the last 3 years, they have published their EEO-1 data, and then included a supplement showing the breakdown of how each individual that picked 2 or more races/ethnicities identified. That way, the data is comparable to other companies but they are also telling a unique and interesting story.

This is just one example of companies not thinking enough about who is reading this report and how the information actually gets used. Some other common poor practices:

  • Including charts with no data labels, so the user has to measure the size of the bar or angle of the pie
  • Reporting on the representation of ‘leaders’ or ‘executives’ but not providing any context—for some companies it could mean the C-suite and another company could mean people managers
  • Changing definitions of how data gets reported from year to year without providing historical data reflecting changes
  • Setting goals/targets for representation of specific groups but not disclosing the baseline figures
  • Only showing your latest year report and not providing links to historical documents

Best-in-class companies provide easy-to-use formats, with granular time-series tables and/or charts and very explicit labels/definitions. Companies should realize these reports are not just an outlet to tell their story and communicate their progress, but tools that stakeholders will use for analysis just like financial reports.

Need human capital data for your planning?