Hybrid working from home improves retention without damaging performance

The study investigates whether hybrid working from home (WFH)—typically two days at home and three days in the office—affects employee retention, satisfaction, productivity, and career outcomes. Post-pandemic, hybrid WFH became widespread among university-educated employees, yet executives have expressed concerns about productivity, innovation, and career development. Prior causal research has largely examined fully remote work in independent-task roles and often found negative productivity effects. Given that most WFH employees globally are on hybrid schedules and work in creative, team-based graduate roles, this study aims to provide causal evidence on the effects of hybrid WFH in such settings.

Existing causal studies focus on fully remote arrangements in roles like call centers, data entry, and helpdesks, often finding negative impacts on productivity and development. However, over 70% of WFH globally is hybrid (commonly three office days, two home days), and most hybrid workers are university graduates engaged in creative, team-based work across science, law, finance, and IT. Concerns about hybrid work include potential harm to innovation and career progression. This study addresses the gap by examining hybrid WFH in a professional, team-oriented context using a randomized control trial (RCT).

Design: A randomized control trial at Trip.com, a large Chinese travel technology multinational, ran from August 2021 to January 2022. Participants: 1,612 graduate employees in the Airfare and IT divisions (395 managers, 1,217 non-managers) covering software engineering, marketing, accounting, and finance. Randomization: Employees with odd-numbered birthdays were assigned to the treatment group, allowed to WFH on Wednesdays and Fridays and work in-office on other days; even-numbered birthdays formed the control group, working in-office five days per week. Baseline surveys collected expectations, backgrounds, and volunteer interest; end-line surveys were conducted on 21 January 2022. Context: During the second half of 2021 in Shanghai, COVID-19 incidence was very low; offices were open, unmasked, and modern open-plan with teams co-located. Outcomes and measures: - Retention/attrition over the six-month experiment. - Job satisfaction and related well-being indices from anonymous surveys (0–10 scales). - Performance: semiannual performance review grades across four periods up to June 2023; detailed subcomponent scores (e.g., innovation, leadership, development, execution); and objective output for 653 computer engineers measured by daily lines of code submitted. - Promotions across four semiannual windows up to June 2023. - Self-assessed productivity expectations before and after the experiment; comparative polling in other divisions post-experiment. Analysis: Differences between treatment and control were assessed with two-tailed t-tests; null equivalence tests were applied to performance grades and lines of code to assess absence of meaningful differences. Subgroup analyses examined heterogeneity by managerial status, commute length, gender, tenure, parental status, engineer role, distance, internet speed, and manager treatment status. Samples decline over time due to attrition from the original cohort.

• Retention and satisfaction: Overall attrition fell by one-third in the hybrid group (mean_control = 7.20%, mean_treat = 4.80%; t(1610)=2.02, P=0.043), a 2.4 percentage point absolute and 33% relative reduction. Job satisfaction improved (mean_control=7.84, mean_treat=8.19; t(1343)=4.17, P<0.001), with higher work-life balance, work satisfaction, life satisfaction, and recommendation to friends, and lower intention to quit. - Heterogeneity in attrition effects: • Non-managers: −3.26 pp (mean_control=8.59%, mean_treat=5.33%; t(1215)=2.23, P=0.026), about 40% reduction. • Managers: no significant change (mean_control=2.96%, mean_treat=3.13%; t(393)=-0.098, P=0.922). • Long commuters: about 52% reduction using above-median split (mean_control=6.00%, mean_treat=2.89%; t(609)=1.87, P=0.062); alternatively, >2 h two-way threshold showed larger drop (mean_control=7.33%, mean_treat=1.89%; t(307)=2.31, P=0.021). • Women: 54% reduction (mean_control=9.2%, mean_treat=4.2%; t(568)=2.40, P=0.017). • Men: insignificant 16% reduction (mean_control=6.15%, mean_treat=5.15%; t(1040)=0.70, P=0.487). - Performance reviews: No significant differences across four semiannual review periods through June 2023: differences in assigned numeric grades (A=5 to D=1) were small and nonsignificant (e.g., Jul–Dec 2021: diff=0.056, P=0.198; Jan–Jun 2022: diff=0.034, P=0.440; Jul–Dec 2022: diff=−0.019, P=0.677; Jan–Jun 2023: diff=0.046, P=0.369). Null equivalence tests supported no meaningful performance impact. - Promotions: No significant differences in promotion rates across four periods (e.g., Jul–Dec 2021: diff=−0.86 pp, P=0.509; Jan–Jun 2022: diff=0.12 pp, P=0.892; Jul–Dec 2022: diff=−0.51 pp, P=0.651; Jan–Jun 2023: diff=−0.99 pp, P=0.328). - Performance subcomponents and objective output: No differences across nine performance subcategories (e.g., innovation, leadership, development, execution). Among 653 engineers, no difference in lines of code submitted between groups. - Beliefs about productivity: Overall self-assessed productivity expectations improved from −0.06% at baseline to +1.48% at end line (t(2658)=−3.84, P<0.001). Managers’ mean expectation shifted from −2.6% to +1.0%, converging toward non-managers (baseline difference managers vs non-managers: −3.28 pp, P<0.001; end-line difference: 0.571 pp, P=0.345). Across four other divisions polled in March 2022 (n=3,461), mean perceived impact was +2.8%, similar to the experimental end line.

The randomized experiment demonstrates that allowing two days per week of WFH in a hybrid schedule substantially improves retention and satisfaction without harming measured performance or promotion prospects in a graduate, team-based tech context. These findings directly address concerns that hybrid work damages productivity or career development: performance reviews, promotion rates, detailed competency scores, and objective coding output showed no detrimental effects. The stronger retention benefits among non-managers, women, and long commuters suggest hybrid policies particularly aid groups likely to value commute savings and flexibility, potentially enhancing diversity and inclusion. Managers’ perceptions shifted from negative to positive after experiencing hybrid work, underscoring the role of evidence and experimentation in updating beliefs. Following the experiment, Trip.com scaled hybrid WFH firmwide, citing substantial cost savings from reduced attrition (approximately US$20,000 per quit), indicating organizational benefits alongside employee welfare gains. Compared with literature documenting drawbacks of fully remote work, these results suggest hybrid arrangements can preserve performance while improving retention and well-being, offering broader societal benefits via reduced commuting and improved work–life balance.

A six-month RCT at a large technology multinational shows that a hybrid schedule with two days per week WFH reduces attrition by about one-third and improves job satisfaction, with no detectable impact on performance reviews, promotions, or objective coding output over up to two years. The effects are particularly strong for non-managers, women, and employees with long commutes. The study contributes causal evidence for hybrid work in creative, team-based graduate roles, countering concerns drawn from fully remote settings. Firms may profitably adopt similar hybrid policies due to retention-driven cost savings and employee amenity value. Future research should test external validity across industries and cultures, explore different hybrid intensities (e.g., 1, 3, or 4 days in office), and examine longer-run effects on innovation, training, and culture.

Generalizability may be limited: the experiment was conducted in a single large Chinese technology firm during a period of low local COVID incidence, with graduate employees in engineering, marketing, accounting, and finance. Findings specifically pertain to a 3-days-in-office/2-days-at-home hybrid schedule and may not extend to more remote configurations (e.g., one or fewer days in-office). Performance measures rely on internal reviews, albeit rigorous, which may not capture all dimensions of productivity and long-term innovation. Sample sizes declined over time due to attrition from the original cohort. The study underscores the importance of full enrollment to avoid potential negative career signaling from volunteering, which could differentially affect subgroups (e.g., women).