Latest / The Diversity Career Podcast with Fexingo: Inclusion, Equity, and Underrepresented Professionals / How a Tech Company Uses Blind Auditions for Internal Promotions
Transcript
- Lucas: We talk a lot on this show about how you get hired. How you get past the first gate, past the resume screen, past the interview loop. But there's another gate that's arguably tougher, and a lot less transparent: the internal promotion. Luna: Right, because once you're in the building, the 'how' of moving up is often this fuzzy mix of who you know, how long you've been there, and whether your manager is an effective advocate. Lucas: Exactly. And the data backs that up. A McKinsey study from 2023 found that women are promoted at a significantly lower rate than men at the first step up to manager, and the gap widens for women of color. The usual explanation is the broken rung. But what if you could just remove the fuzz? Remove the manager sponsorship variable entirely? Luna: That's what one company I've been looking at decided to try. It's a mid-sized tech firm called DataNova — about 800 employees, mostly engineering and product roles. And they implemented something they call blind auditions for all internal promotions above the individual contributor level. Lucas: Blind auditions — like an orchestra. Candidates perform, the evaluators don't see the performer. How does that translate to a promotion process? Luna: So, when a manager wants to promote someone from, say, senior engineer to staff engineer, the candidate doesn't just get a nod from their boss. Instead, they submit two things: a work sample — a piece of code, a strategy doc, a project retrospective — and a recorded presentation answering a set of standardized questions. Both are stripped of the candidate's name, tenure, team, demographic markers. Then a rotating panel of three to five people from outside the candidate's chain of command scores them on a rubric. Lucas: And the manager's nomination itself — does that carry weight in the scoring, or is it just what gets the candidate into the pipeline? Luna: The nomination triggers the process, but the manager has no say in the score. The panel doesn't even know who nominated them. The candidate's identity is revealed only after the score is finalized, and the promotion is approved if the score meets a threshold set by the people team. No exceptions. Lucas: That's a radical break from how most promotions work. In a typical company, the manager is the primary advocate. They argue for their person in a calibration meeting. Here, the manager is essentially removed from the evaluation. How did that land? Luna: It landed with a lot of friction, especially in the first cycle. Managers who had been used to 'sponsoring' their direct reports felt disempowered. They said things like, 'I know my people best, I should have a say.' And there was some truth to that — but the company's argument was that manager sponsorship is exactly where bias creeps in. A 2021 Harvard Business Review study showed that managers tend to sponsor employees who are similar to themselves, demographically and in terms of working style. So the system was explicitly designed to break that. Lucas: And the results after, say, two cycles? Luna: DataNova ran the blind audition process for two promotion cycles, covering about 120 candidates. They shared some internal numbers with me. Promotions of women and underrepresented groups increased by 34% compared to the previous two cycles. But here's the part I found even more interesting: overall promotion rates stayed roughly flat. So it wasn't that more people got promoted — it was that the mix shifted. The people being promoted were more diverse, and the work sample scores showed that the candidates who were previously overlooked had comparable or better scores than those who had been promoted through the old system. Lucas: That's the key test, isn't it? You're not lowering the bar. You're just removing a filter that was letting some people through and blocking others who were equally qualified. The question is whether the bar itself is valid. What did the rubric look like? Luna: The rubric had four dimensions: technical depth, strategic thinking, collaboration, and communication. Each scored 1 to 5. The panel members had to provide written justification for each score. The company did a calibration exercise before each cycle where all potential panelists scored a sample submission together and discussed discrepancies to align on standards. They also rotated panels so no one served on back to back cycles. Lucas: That calibration piece is huge. Without it, you just get individual subjectivity dressed up as objectivity. But with calibration, you're training evaluators to see the work, not the person. Did DataNova track retention of the blind-promoted cohort? Luna: They did. Retention at 12 months post-promotion was 22% higher for the blind audition cohort compared to the historically promoted cohort. The hypothesis is that because the promotion was based on demonstrated skill, not internal politics, the promoted employees felt more validated and less likely to leave for a role elsewhere. Also, they reported higher trust in the fairness of the process in engagement surveys. Lucas: That retention number might be the most compelling argument for this. Promotions are expensive. When you promote someone and they leave within a year, you've lost the investment in their development and the institutional knowledge. If blind auditions reduce that churn, the ROI is clear. Luna: Exactly. But it's not a complete solution. DataNova still saw some resistance. A few managers left the company because they felt their role as developer of talent was undermined. And there are roles where a work sample is hard to define — like a product manager whose impact is spread across many projects over a quarter. DataNova ended up allowing a portfolio submission with a narrative for those roles, but they still anonymized it. Lucas: That seems like a reasonable adaptation. What about the panelists themselves? Did they get fatigued? Scoring 60 submissions per cycle is a lot of work. Luna: It was a significant time commitment. Each submission took about 30 to 45 minutes to evaluate. The company compensated panelists with a small bonus and made it a recognized contribution in performance reviews. They also capped the number of submissions any one panelist could evaluate at ten per cycle to prevent burnout. And they noticed that panelists who served multiple cycles got faster and more consistent. Lucas: So the system improves with use. That's encouraging. But I want to push on something. The blind audition model works well when you have a clear, discrete role with a defined scope. What about promotions to director or VP level, where the role is more about influence and cross-functional leadership? Can you really evaluate that from a work sample and a recorded presentation? Luna: DataNova hasn't applied it to director level yet. They're starting with individual contributor and first-level manager promotions. For director roles, they still use a traditional panel interview with stakeholders. But they're experimenting with a modified version where the candidate's strategic presentation is anonymized, and the panel doesn't know their background until after scoring. Early feedback is mixed — some stakeholders feel they need to know the person's history to assess their fit. Lucas: Which is exactly where bias creeps back in. 'Fit' is often code for 'someone I can see myself working with.' And that tends to favor people who are like the existing leadership. It's the mirror test. Luna: The mirror test — I love that phrase. And DataNova's CEO actually used that term when they announced the blind audition policy. She said, 'We've been promoting in our own image for too long. This is how we stop.' Lucas: That's a strong internal message. And it's backed up by data from outside the company. There's research from the National Bureau of Economic Research on blind auditions for orchestras—when they went blind, the probability of a woman advancing from the preliminary round increased by 50%. And the same logic applies here: when you focus on the output, you see talent you otherwise miss. Luna: Right. And that's really the core of this episode. DataNova's experience shows that it's possible to redesign a promotion system to be more equitable without sacrificing quality. It takes work, it creates friction, but the outcomes — more diverse promotions, higher retention — are worth it. Lucas: It also requires a culture that's willing to challenge the assumption that managers know best. That's not easy in a lot of organizations. But DataNova's data suggests that the blind audition model can be a powerful tool. I think we should check back in a year and see how it's scaling. Luna: Absolutely. And speaking of value, you know, we spend a lot of time on this show digging into specific examples like this one. We do it deliberately without advertising or sponsors. If you find these conversations helpful, one way to support that choice is through a small contribution at buy me a coffee dot com slash fexingo. It keeps us ad-free and focused on the details. Lucas: Yeah, listener support is what makes this possible. Every little bit helps us keep bringing you these concrete cases. Now, back to DataNova — one more thing I wanted to ask. Did they see any unintended consequences? Like, did people start gaming the system? Luna: Good question. There was some early concern that candidates would spend too much time polishing their submission instead of doing their actual job. But the company set a strict page limit and required the work sample to be something the candidate had produced in the normal course of work, not a special project. That helped. And the panelists were trained to look for genuine work, not performance. Lucas: That's smart. You don't want the process to become a separate performance. You want it to reflect the actual work. So the system seems to pass the sniff test. The question is whether it can survive leadership changes. If a new CEO comes in who believes in sponsorship, the whole thing could be scrapped. Luna: That's the biggest risk. The current CEO is the champion. She's staked her reputation on it. But she's also been transparent about the fact that the process is a pilot. It's not enshrined in the company handbook. So it's fragile. But she's hoping the data will speak for itself and make it hard to reverse. Lucas: Data-driven persistence. That's about as good a strategy as you can have in the diversity space. For now, DataNova offers a rare example of a company that actually redesigned its promotion system from the ground up, measured the results, and shared them. That's more than most organizations do. Luna: Yeah. And I think it's a model that could be adapted, even in part, by other companies. Maybe not full blind auditions, but anonymized work samples, or calibrated panels, or limiting manager influence. Each piece can help reduce bias. The key is to start. Lucas: Start small, measure, iterate. That's good advice for any DEI initiative. Thanks for bringing this one in, Luna. Luna: Thanks for digging into it with me. And thanks to our listeners for sticking with us. If you have an example of a company doing something interesting in the inclusion space, let us know. We love hearing from you.