Background: Effective strategies are needed to address obesity, particularly among minority and low-income individuals. Simple point-of-purchase interventions have the potential to improve nutrition choices of a diverse population. Methods: We tested the effectiveness of a 2-phase food labeling and choice architecture intervention among different racial and socioeconomic groups in 4,642 employees who were regular patrons at a large hospital cafeteria in Boston, MA. After collecting baseline sales data for 3 months, we implemented Phase 1 to label all items green (healthy), yellow (less healthy), or red (unhealthy). In Phase 2, we manipulated “choice architecture,” making green-labeled items more accessible and red-labeled items less accessible. We calculated the proportion of items purchased by an employee labeled green or red during each 3-month period (baseline, Phase 1, Phase 2) and assessed person-level changes. We linked employee cash register data to human resources files to determine demographics, including race/ethnicity (self-reported) and job type (measure of socioeconomic status). We tested the interaction between intervention phase and race or job type to determine differences in intervention effects across subgroups. We calculated average calories and price per beverage during each phase. Results: Employees were white (73%), black (10%), Latino (7%), and Asian (10%). At baseline, black and Latino employees purchased higher proportions of red items compared to white employees (33%, 28%, and 18%, respectively, p<0.001) and lower proportions of green (33%, 38%, and 48%, p<0.001). Purchases by Asian employees did not differ from white employees. Adjusting for demographics, labeling decreased all employees’ purchases of red items (-11.2%, 95% CI -13.6% to -8.9%) and increased green purchases (6.6%, 95% CI 5.2% to 7.9%). The greatest effect was for red beverages (-23.8%, 95% CI -28.1 to -19.6). Choice architecture further decreased red purchases after labeling. Intervention effects were similar across all race and job types (p>0.05 for tests of interaction between race or job type and intervention phase). At baseline, average calories per beverage purchased by black and Latino employees was higher than for white employees (126, 113, and 87 calories, respectively, p<0.001). During the 2-phase intervention, mean calories per beverage decreased similarly for all racial groups and job types (-15 calories, 95% CI -18 to -13), but the mean cost per beverage did not change. Conclusions: Despite racial and socioeconomic differences in purchasing of healthy foods at baseline, a simple color-coded labeling and choice architecture intervention improved food and beverage choices among employees from all backgrounds.