Over the weekend, Fast’s CEO Domm Holland posted a short Twitter thread about growing from 2 to 120+ people in 18 months while maintaining “an exceptionally high talent bar” and offered some tips based on Fast’s hiring process. But as I was reading the thread, I was struck by how many of the practices seemed likely to perpetuate a monoculture. So I drafted this post to offer a counterpoint to some of his recommendations by surfacing potential alternatives.
Note that I am not responding to all of Holland’s tips: “treat your people well” is grounded in strong evidence for supporting a diverse and inclusive workplace that leads to high performance. I’m also not responding specifically to Fast’s culture, although the thread responses do point out potential issues across a variety of domains, from rescinding offers to using Nigerian devs at very low wages to build V1 of the product and then terminating them without cause.
Before going through the tips, it is important to remember that hiring practices should be evidence-based and so I needed to look at Fast’s hiring data before responding to Holland’s thread; maybe their practices are a secret recipe for diversity and I’m simply wrong about their monocultural nature. Since Fast doesn’t have a public diversity report that I could find, I gathered 110 people on LinkedIn who identify themselves as working at Fast. I then coded for perceived ethnicity/gender (using names and Twitter/other photos if they didn’t have one on LinkedIn), had another blind rater do the same, and compared; there was only one person we coded differently, so I removed them from the sample, leaving 109.
Obviously, this methodology leaves out other important forms of diversity and coding gender/ethnicity using LinkedIn photos and names is flawed; ideally, all companies would release their self-reported diversity data to avoid these limitations.
To understand Fast’s diversity data, I compared it to Google’s hiring data from their 2020 diversity report. Since Holland specifically calls out growth in the last 18 months and focusing on “the top 1% of major tech companies”, Google feels like an appropriate benchmark.
Fast | ||
Male | 68.8% | 67.5% |
Female | 31.2% | 32.5% |
White | 62.4% | 43.1% |
Asian | 28.4% | 48.5% |
Latinx | 4.6% | 6.6% |
Black | 4.6% | 5.5% |
Given that Fast achieved essentially identical gender diversity and significantly less ethnic diversity than Google, which has itself recently paid fines for documented hiring biases, a reply to Holland’s thread seems justified by the data.
Holland Tip #1: “We strictly do all recruiting internally” and “recruiting internally keeps tighter quality control”
Commentary: It isn’t entirely clear what Holland means here, since recruiting can include any number of responsibilities, from sourcing and screening to interviewing and negotiation. There are entirely valid reasons to keep any and all of those processes internal but unless specifically designed to increase diversity, internal processes tend to favor the status quo. And since the status quo in tech is overwhelmingly white and male and privileged, that means perpetuating those characteristics.
Alternative Tip #1: Because diversity begets diversity, use outside resources (including diverse external sourcing that you pay for) to challenge the status quo when needed.
Holland Tip #2: “we receive A LOT of inbound” and “yet still put 90% of effort into outbound sourcing to find the exact background & skill set we are looking for”
Commentary: A wide funnel of inbound is a great sign of traction but not always useful for increasing diversity, since those likely to apply are those likely to know, and those likely to know are those likely to already be in your network. So outbound is an absolutely critical part of increasing diversity. The red flag here is how that outbound is happening: not to broaden the inbound pipeline but to narrow it. “Exact background & skill set” are often code words for biased filters like specific universities or companies that already have issues with bias; just as predators aggregate pollutants from the animals they eat, relying on biased sources means you aggregate those biases.
Alternative Tip #2: Keep a wide funnel in your inbound by using clear job descriptions with bias reducing features and use outbound in a targeted way to widen, not narrow, wherever you have measured diversity issues.
Holland Tip #3: “we almost exclusively hired experienced people”
Commentary: “Experienced” is an interesting euphemism but Holland fortunately defines his usage as “hiring people who are currently thriving in a place that would be our next level up.” This seems at odds with his first tip, since hiring specifically from competitors is very much like using outside recruiters (you are relying on the filtering of others) but as with Tip 2, the net effect is that you aggregate their biases.
Alternative Tip #3: Monitor source diversity to avoid overindexing on single sources, be they specific schools, companies, or industries, and target increasing source diversity independently to achieving candidate diversity.
Holland Tip #4: “a-players attract a-players” and “the more we have focused on the best people, the higher the quality of applicant we get”
Commentary: The tendency of like to attract like (often called homophile) is well documented and when used purposefully, it can actually be a diversity superpower: diverse companies tend to get more diverse because of it. But when type matching (like “a-players”), homophile generally reduces diversity by creating monolithic patterns for what is considered “the best”.
Alternative Tip #4: Recognize and plan for broad representations of talent by having clear plans for recruiting unique skills and talents, then valuing and utilizing them. Think of Venn diagrams that touch but don’t overlap more than 50%.
Holland Tip #5: “our team are the hiring panel, most of our team interview, screen and make hiring decisions”
Commentary: Unless a hiring manager is specifically attuned to increasing the diversity of their team, using hiring panels (rather than single hiring managers) increases diversity because it increases the number of potential advocates for underrepresented talent. But this only occurs if the panel itself is diverse (again, homophile applies), trained and attuned to increasing diversity, and when unanimous decisions are not required.
Alternative Tip #5: Use diverse panels with specific training and allow for non-unanimous decisions; augment with external panel members when there is not enough internal diversity or availability.
Holland Tip #6: “internal referrals are exceptionally value [sic], they know the best people they have worked with”
Commentary: Internal referrals are specifically homophilic; as Holland says, we know who we have previously worked with. This leads to the compounding of problems/virtues: diverse teams get more diverse, monocultures get more monocultural.
Alternative Tip #6: Be clear about the value of diverse referrals by clearly publishing diversity data and encouraging diverse referrals both individually and systematically (either through rewards or refusal to consider overrepresented referrals).
Hiring high performing teams is hard and so for founders that believe in doing hard work, Holland’s thread is a seductive opus to that effort. But as a white male founder, it is easy to think that easy things are hard simply because we believe they should be, even when the easiest thing to do is hire other white males: they are probably who you know, who you’ve worked with, and who are easiest to attract.
So when we talk about doing the work, we need to recognize what the real work is. Literal decades of research at this point tell us how important diversity and inclusion are to high performing teams; a failure to explicitly address those challenges in a thread about hiring practices is to acknowledge that you either don’t understand the current state of diversity in the workplace or aren’t actually committed to team as a strength. Either way, we can and should do better; as Holland says at the end of his thread, “people deserve it”.
Side Note: I recently read some of my older blog posts and was aghast as the privilege encoded in them (along with the really terrible advice resulted). While Holland’s thread created a cacophony of justifiable anger directed at both him and Fast, I wonder if that facilitates individual change or mires it in defensiveness; could I look back at my entries and acknowledge their flaws had I been attacked for them at the time? Also, I wonder at the shelf life of even the alternative tips I list here; hopefully new science proves some of these suboptimal and we get ever better at creating the more equitable world we want.