This article summarises a peer-reviewed field study published in the Journal of Economic Behavior & Organization (2025), co-authored by Iris Bohnet, Oliver P. Hauser, and Ariella S. Kristal of Harvard Kennedy School. The study investigates whether gender and race disparities in performance appraisals stem from employees' self-ratings, managers' evaluations, or the interaction between the two. Leveraging a natural experiment at a multinational financial services firm — a 2016 software glitch that prevented managers from viewing self-evaluations — the researchers analysed four review cycles (2015–2018) to isolate the causal effect of self-rating visibility on final performance scores. Key findings indicate that women, particularly women of colour, consistently assigned themselves lower self-ratings; managers rated people of colour lower regardless of self-rating visibility; and concealing self-ratings reduced anchoring effects but did not eliminate demographic disparities because managers substituted historical ratings as an alternative anchor. Notably, newcomers with no rating history who also had hidden self-ratings showed reduced gaps for women of colour. The article concludes that addressing performance rating bias requires both process-level interventions (withholding self-ratings) and manager-level interventions (structured criteria, bias interrupters, and demographic auditing). Key insights: Women, and especially women of colour, consistently self-rate lower than men — a pattern attributed to social norms and anticipated backlash against self-promotion rather than actual performance differences. Race-based penalties in manager ratings persisted even when self-evaluations were hidden, indicating that managerial bias operates independently of self-rating anchoring and is not resolved by process design changes alone. When self-ratings were hidden and no historical rating data existed (newcomers), women of colour achieved parity with white women and men — suggesting that anchoring on past scores is a secondary mechanism that perpetuates inequity. Practical takeaways: Withholding self-evaluations from managers until after initial ratings are submitted can reduce anchoring effects, with the strongest benefit observed for employees without prior rating histories. Because race gaps in manager ratings persist independently of self-rating visibility, demographic outcome auditing and structured calibration processes represent a distinct and necessary layer of intervention beyond process sequencing changes.