The Challenge
SpectrumTech, a 200-person B2B SaaS company, had a diversity problem they could not solve with good intentions alone. Despite an explicit commitment to diverse hiring, their engineering team was 82% from the same demographic. Three years of diversity initiatives had produced marginal improvements (18% to 22% diverse hires).
The problem was not a lack of effort. SpectrumTech had diversity training, inclusive job descriptions, and diverse interview panels. But their outcomes did not change because the underlying process had structural issues they could not see.
"We were doing all the 'right things' on the surface - diverse panels, inclusive language, ERGs. But our pipeline was still narrow because we were sourcing from the same places and screening the same way."
The Diagnosis
WorkSwipe's audit of SpectrumTech's hiring data revealed three structural issues:
- Sourcing concentration - 78% of candidates came from 3 sources (LinkedIn, employee referrals from existing network, one university pipeline). All three had limited diversity
- Resume screening bias - Analysis showed candidates from non-traditional backgrounds (bootcamps, self-taught, career changers) were screened out at 3x the rate of traditional CS degree holders, despite equal technical performance when they reached interviews
- Assessment format bias - Whiteboard coding interviews favored candidates with competitive programming backgrounds, which correlated with demographics, not job performance
The Solution
Step 1: Diversify sourcing. WorkSwipe expanded candidate sourcing to 12+ channels including HBCUs, coding bootcamp alumni networks, professional organizations for underrepresented groups, and open-source communities. AI matching evaluated candidates on demonstrated skills, not pedigree.
Step 2: Skills-based screening. Replaced resume-first screening with skills assessment. Candidates were evaluated on what they could do (via work samples and technical discussions), not where they learned it.
Step 3: Structured interviews. Replaced unstructured conversations with standardized rubrics. Every candidate answered the same questions, evaluated on the same criteria, scored on the same scale. Reduced interviewer discretion while maintaining technical rigor.
Step 4: Bias monitoring. WorkSwipe's analytics tracked pass-through rates by demographic at each pipeline stage. Any stage showing statistically significant differential was flagged for review.
Results (12-Month Comparison)
- Diverse hires: 45% (up from 18% - a 2.5x increase)
- Sourcing channels: 12+ (up from 3)
- Technical bar maintained - Code review scores and 90-day performance ratings were statistically identical across all demographic groups
- Time-to-hire unchanged - Process changes did not slow hiring (24 days average)
- Employee satisfaction: +18 points on inclusion survey
- Referral diversity: +35% - As the team diversified, referral pipelines naturally broadened
"The key insight was that our bar was not too high - our filter was too narrow. When we evaluated skills instead of credentials, we found exceptional candidates we had been systematically overlooking."