Renée Carignan
Optemization1. Can you share case studies or examples where your DEI feature has been successfully implemented and the outcomes it achieved?
Twilio debugs its diversity hiring funnel with Gem
Gusto’s talent team partners use Gem to become even more strategic hiring partners
2. How do you measure the effectiveness and accuracy of your DEI suggestions? What metrics or KPIs do you track?
The inference model we use is from Namsor, a company created by academic researchers in Europe. Namsor is one of only a couple trusted companies who provide algorithmic inference on gender and race/ethnicity. They are academics first and have been cited in ~100 academic papers in 2023 alone.
As an example, here’s one research paper from 2022 that uses Namsor’s data.
3. How frequently is your DEI feature updated to reflect the latest research and best practices in the field?
Gem’s race/ethnicity inference comes from the aforementioned 3rd party provider, Namsor, who is continually updating their model based on their research and newer data sets. In terms of the DEI functionality that’s available in Gem, our product team is also continuously fielding feedback from customers and improving the functionality we have. For example, in Talent Compass — Gem’s reporting solution — we recently shipped the ability to breakdown demographic data between self ID data imported from the ATS and inferred data. Another feature we recently shipped is for customers to create custom groupings to enable reporting on specific demographic groups they care about; e.g. within an organization, the definition of diversity in engineering may include women and BIPOC, but in G&A or marketing, diversity may be defined as men and Asians. Custom groupings allow you to configure your diversity reports in a way that makes sense to your company and its goals.
4. Are there any third-party audits or certifications that validate the effectiveness and accuracy of your DEI feature?
Namsor cites various audits and studies, including but not limited to the below:
- Elsevier audit
- Uber benchmark
- Harvard/Chicago University study
5. Can you explain the privacy and data security measures in place to protect sensitive information used by the DEI feature?
By default, DEI predictions are currently viewable only in aggregate in reports when a minimum sample size is reached so that false conclusions aren’t made from small datasets, by ensuring that:
- A minimum sample size (n=20) exists before disclosing the gender of a given candidate group
- A minimum sample size (n=5+) exists before disclosing the race/ethnicity of a given candidate group
Team admins can enable Full Permissions on a per-user basis to show gender and/or race/ethnicity at the individual profile-level.
And finally, we also offer the ability to disable the gender and race/ethnicity inference on candidates based on their geography. This can be helpful when gender or race/ethnicity is not a factor in hiring abroad, and having that data for those candidates is therefore not needed. Admins on your team can configure these location-based rules to ensure that this data is available only when needed.
6. What is the roadmap for future enhancements to the DEI feature, and how do you plan to continue evolving its capabilities?
DEI has always been top of mind for Gem, and a core value of our company. To continue building upon our DEI capabilities, we have a few investment areas top of mind, including:
- Incorporating DEI into additional relevant product surfaces in Gem to enhance key workflows, like Application Review and Prospect Search
- Broadening the race/ethnicity and gender categories to reflect the diversity talent teams want insight into: currently, our race/ethnicity categories mirror that of the standard EEO groupings that the US Federal Government uses, which are incredibly broad and ultimately non-inclusive groups. We find that many forward-thinking talent organizations crave greater accuracy and granularity of the race/ethnicity and gender groupings we provide.
- Building self-ID earlier into the process: currently, employers collect EEO data from candidates, but that doesn’t happen until the application process. As most talent teams know, much of the hard work and effort required to hire a candidate is done before they even apply through cold outreach, so finding ways to capture diversity higher up in the funnel is extremely valuable to inform outreach & attraction strategy.
Keep in mind that our roadmap is deeply informed by customer needs and that you always have an open door to our head of product and broader product team, so please share feedback to help us prioritize the right investments for DEI.