Hybrid working from home improves retention without damaging performance

Once the experiment ended, the Trip.com executive committee examined the data and voted to extend the hybrid WFH policy to all employees in all divisions of the company with immediate effect. Their logic was that each quit cost the company approximately US$20,000 in recruitment and training, so a one-third reduction in attrition for the firm would generate millions of dollars in savings. This was publicly announced on 14 February 2022, with wide coverage in the Chinese media. Since then, other Chinese tech firms have adopted similar hybrid policies.

This highlights how, contrary to the previous causal research focused on fully remote work, which found mostly negative effects on productivity, hybrid remote work can leave performance unchanged. This suggests that hybrid working can be profitably adopted by organizations, given its effect on reducing attrition, which is estimated to cost about 50% of an individual’s annual salary for graduate employees. Hybrid working also offers large gains for society by providing a valuable amenity (perk) to employees, reducing commuting and easing child-care.

The experiment was conducted in a Chinese technology firm based in Shanghai. Although it might not be possible to replicate these results perfectly in other situations, Trip.com is a large multinational firm with global suppliers, customers and investors. Its offices are modern buildings that look similar to those in many American, Asian and European cities. Trip employees worked 8.6 h per day on average, close to the 8 h per day that is usual for US graduate employees. The business had a large drop in revenue in 2020 (see Extended Data Fig. 4), followed by roughly flat revenues through the 2021 experiment period into 2022, so this was not a period of exceptionally fast or slow growth. As such, we believe that these results— that is, the finding that allowing employees to WFH two days per week reduces quit rates and has a limited effect on performance—would probably extend to other organizations. Also, this experiment analysed the effects of working three days per week in the office and two days per week from home. So, our findings might not replicate to all other hybrid work arrangements, but we believe that they could extend to other hybrid settings with a similar number of days in the office, such as two or four days a week. We are not sure whether the results would extend to more remote settings such as one day a week (or less) in the office, owing to potential challenges around training, innovating and culture in fully remote settings.

Finally, we should point out two implications of the experimental design. First, full enrolment into hybrid schemes is important because of concerns that volunteering might be seen as a negative signal about career ambitions. The low volunteer rate among female employees, despite their high implied value (from the large reductions in quit rates observed), is particularly notable in this regard. Second, there is value in experimentation. Before the experiment, managers were net-negative in their views on the productivity impact of hybrid working, but after the experiment, their views became net-positive. This highlights the benefits of experimentation for firms to evaluate new working practices and technologies.