I’ve been interested in diversity sourcing for many years now. I’ve helped build sourcing teams and have created sourcing strategies designed to address diversity challenges at several companies, including Fannie Mae, Starbucks, and Apple. Today, we have substantial evidence that companies with diversity initiatives are more likely to outperform those without one. This is not surprising to me since the customer demographics in the US are becoming more diverse, and so diversity reach in the product, marketing, and sales teams will help drive success. For recruiting, this means an increasing need for pipelines of diverse, qualified talent.
While business leaders are rightly concerned with hiring, promoting and retaining diverse professionals, it falls on recruiting teams to find and attract this talent. And whenever there is a challenge that combines marketplace scarcity and hard-to-find skill-sets, recruiting teams turn to their expert sourcers who bring their unique abilities and tools to help boost the pipeline.
Since we don’t live in an ideal world and many recruiting teams don’t have expert sourcers to turn to for help, they often have to rely on tools to help uncover diverse talent. And that is what I’m writing about today – sourcing tools that focus on diversity. These can be used by anyone, even someone who is not an expert sourcer, and they can even help the expert sourcer achieve better results faster.
DIFFERENT APPROACHES TO FINDING DIVERSITY PROFILES
A conventional approach to finding diverse profiles, if you’re an expert sourcer, is to construct LONG Boolean queries, with lists of 100+ female first names, or Hispanic first and last names, or sororities, associations and fraternities associated with African Americans to generate diverse candidates. Glen Cathey wrote the classic blog-post on this topic comparing various query-based approaches to get diverse candidates on LinkedIn, and it remains a great resource today for experts. While this approach certainly brings back many diverse candidates, it also runs into the limitations imposed by search engines. For example, Google limits search queries to 32 words and ignores subsequent words (so at most 32 names can be used in the search), and similar constraints exist for LinkedIn, especially for those who don’t have a full Recruiter license.
The alternative to a query-based search is the index-time approach. Here, when a tool has its own database of profiles and its own search engine on top, the tool can use its own criteria to build diversity filters at the time of indexing (like Google does when indexing pages on the web for your searches). For example, it can use tens or hundreds of thousands of Hispanic first and last names from census data to power the Hispanic filter. It can use country of an undergraduate degree and sororities and fraternities for diversity classification. It’s even possible to use profile images to help in classification.
The third approach is where users themselves enter their gender or ethnicity. Facebook, for example, makes it easy to search for women profiles, as profiles already have a gender associated with them. Facebook is not as helpful for other diversity classes, and they are not the top place for professional information.
I don’t think any approach is perfect, but the goal should be to increase the chances that we find diverse profiles that fit our search query. All approaches need to balance between false-positives (e.g., profiles classified as female when they are actually not), and false-negatives (e.g., profiles that are actually female not being classified as female). My preference is for leaning towards false-positives, as it just means that as a recruiter I need to sift through a few extra profiles. On the other hand, if I used a query with just the 100 most popular women’s names, I am biasing towards false-negatives as lots of women who do not have a popular name will not show up in my search.
For this blog post I decided to focus on three tools I had access to that use an index-time approach to diversity filters – Entelo, SeekOut, and Hiring Solved. None of the other tools I had access to (LinkedIn or Hiretual) supported index-time diversity filters. I welcome suggestions and access to other tools that have this ability to expand the comparison for a future blog post.
I decided to perform an apples-to-apples comparison using some defined queries. The three I used were:
- Query-1: Profiles with current title “Data Scientist” at Facebook in the US.
- Query-2: Profiles with current title “Developer OR Engineer” at Microsoft in the US.
- Query-3: Profiles with current title “District Manager” at Starbucks in the US.
I include HiringSolved here to make mention of their diversity search/filter functionality. They are different than others in their approach. Unlike Entelo and SeekOut who filter out non-diverse profiles in their results, HiringSolved has “Diversity Boost” filters which prioritize diverse profiles to the top of the search results. This is a great sourcing tool with so much powerful functionality, but I was not as impressed with their diversity search results. I found far fewer results than the other two tools I was testing, so I decided to leave HiringSolved out from a detailed comparison. They get accolades from me for including diversity filters and also for being creative and unique in how they deliver results. We are lucky as an industry to have them in our arsenal of tools to choose from. PS: Other people may have different results with HiringSolved diversity filters, there are many kinds of profiles to search for, and some searches may produce stronger results than those I used.
Entelo and SeekOut Diversity Filter Results
The table below shows the diversity results across the three queries for Entelo and SeekOut.
|Query-1: Facebook||Query-2: Microsoft||Query-3: Starbucks|
|Female||65 (20%)||84 (25%)||2,500 (11%)||3,238 (13%)||449 (45%)||504 (49%)|
|Hispanic||5 (0.2%)||1 (0%)||894 (3.8%)||1,169 (4.8%)||82 (8.2%)||99 (9.5%)|
|African American||0 (0%)||1 (0%)||215 (0.9%)||45 (0.2%)||20 (2%)||0 (0%)|
|Veteran||1 (0%)||0 (0%)||411 (1.7%)||64 (0.3%)||18 (1.8%)||0 (0%)|
By carefully scoping the queries I have roughly the same # of total profiles showing up in the top row. This allows a more apples-to-apples comparison. Both Entelo and SeekOut seem to perform comparably, with Entelo doing slightly better on women and Hispanic filters, and SeekOut doing better on African-American and veteran filters. I did a spot check the results produced by the filters, and for the most part the quality of results was outstanding.
The bigger picture benefit is an even more important takeaway. For example, for Query 3 on Entelo and SeekOut, a recruiter will have to flip through half as many pages as LinkedIn to find the same # of women, roughly 10-times fewer pages to find the same # of Hispanics, and 50+ times fewer pages to find the same # of African-Americans and Hispanic profiles. Time back in your day.
Entelo and SeekOut Diversity User Interface
Both Entelo and SeekOut provided simple and easy to use interfaces for diversity that could be used equally well by an experienced sourcer or a junior recruiter. I’m saying this because I know from experience that ease-of-use is directly correlated with user adoption. This may not matter to all who are reading this, but anyone who has needed to push a recruiting technology to large dispersed recruiting teams will appreciate when the tools are easy to use. My personal view is that with access to tools like Entelo and SeekOut, most recruiters will feel inclined to include diversity candidates in slates they submit to hiring managers.
Figure 1: Entelo user interface for diversity. You can select four major categories and also use them in combinations, e.g., Black AND Female. They also support filtering by various networks and groups.
Figure 2: SeekOut user interface for diversity. You can select major categories of diversity, and see filtered results with one click. One aspect that stood out for me was that SeekOut shows how many profiles correspond to a diversity category in advance of the recruiter applying that filter.
Addressing Selection Bias
In addition to diversity sourcing and pipeline building, more and more companies are taking steps to address selection bias in recruiting. This subject alone could be a blog post in itself. Both Entelo and SeekOut allow for users to present (to hiring managers) candidates that meet skills and experience criteria but hide obvious clues to race or gender by eliminating names and profile pictures. This may seem way out there to many reading this, but companies are beginning to challenge themselves by asking: “are we taking reasonable steps to eliminate bias in our processes?”
In my two decades in the recruiting business I have wondered on many occasions (wondered out loud or just in my head) “What would happen if I put ten perfect resumes in front of hiring manager X without any name on them” or “What if I could send ten on-target profiles with no pictures at all”. If you’ve been in the business as long as I have, you have come face to face with the baggage that decision makers bring to selecting people to interview. That is why these two tools are providing us welcome functionality that we can use or not.
When I recruited in India, candidates wanted me to know their names so that I would select them because of their father’s name – they wanted me to be biased in favor of them because of their name. Today, here and now, ask yourself this: what real value is learned from a name or a person’s profile picture? If you feel that important lessons can be learned by focusing on what is truly needed to do the job, then you might appreciate what Entelo and SeekOut are doing.
With Entelo go to Your Settings, then Search and then choose Anonymize search results. Profile pictures are gone, and names turn to initials.
With SeekOut it’s front, and center just choose Blind Hiring Mode. All profile pictures turn to the same cat image (no bias from people’s cat preferences) and names to just initials.
I have always had a passion for sourcing for diversity, so this was a fun exercise for me. In particular, I had never thought deeply about query-time versus index-time tradeoffs, and limits associated with query-time searches.
My second big takeaway is that new tools are being built, SeekOut and Entelo being two examples, that make it easy for us to source diverse candidates without having to become Boolean experts. We have always advanced by building and using tools that empower us, so it’s exciting to try these.
I also loved the way that SeekOut and Entelo are taking steps to mitigate bias when sharing results with hiring managers. Admittedly, this is one of many steps that need to be taken, but again a great start. I am proud to be in an industry that takes diversity sourcing seriously and that supports innovative technologies like HiringSolved, Entelo, and SeekOut to drive better results. I encourage all of you to try out these new tools and experience the power for yourselves.