Hybrid working from home improves retention without damaging performance

Another key question for Trip.com was the effect of hybrid WFH on employee performance. To assess that, we examined four measures of performance: six-monthly performance reviews and promotion outcomes for up to two years after the start of the experiment, detailed performance evaluations, and the lines of code written by the computer engineers. We also collected self-assessed productivity effects of hybrid working from experimental participants before and after the experiment to evaluate employee perceptions.

Performance reviews are important within Trip.com as they determine employees’ pay and career progression, so are carefully conducted. The review process for each employee is built on formal assessments provided by their managers, co-workers, direct reports and, if appropriate, customers. They are reviewed by employees, collated by managers and by the human resources team, and then discussed between the manager and the employee. This lengthy process takes several weeks, providing a well-grounded measure of employee performance. Although these reviews are not perfect, given their tight link to pay and career development, both managers and employees put a large amount of effort into making these informative measures of performance.

Figure 3 reports the distribution of performance grades for treatment and control employees for the four half-year periods: July to December 2021, January to June 2022, July to December 2022 and January to June 2023. These four performance reviews span a two-year period from the start of the experimental period. Across all review periods, we found no difference in reviews between the treatment and control groups (Extended Data Table 4 and ‘Null results’ section of the Methods).

Figure 4 reports the distribution of promotion outcomes for the treatment and control employees for the same periods. We see no evidence of a difference in promotion rates across treatment and control employees. This is an important result given the evidence that fully remote working can damage employee development and promotions.

We also analysed the effects of treatment on performance grades and promotions for a variety of subgroups, including managers, employees with a manager in the treatment group, longer-tenured employees, longer-commuting employees, women, employees with children, computer engineers and those living further away, as well as looking at whether internet speed had any effect. We found no evidence of a difference in response to treatment across these groups (Extended Data Table 5).

The experiment also analysed two other measures of employee performance. First, the performance reviews at Trip.com have subcomponents for individual activities such as ‘innovation’, ‘leadership’, ‘development’ and ‘execution’ (nine categories in all) when these are important for an individual employee’s role. We collected these data and analysed these scores for the four six-month performance review periods. We found no evidence of a difference across these nine major categories over the four performance review periods (Extended Data Table 6). This indicates that for categories that involve softer skills or more team-focused activities—such as development and innovation—there is no evidence for a material effect of being randomized into the hybrid WFH treatment. Second, for the 653 computer engineers, we obtained data on the lines of code uploaded by each engineer each day. For this ‘lines of code submitted’ measure, we found no difference between employees in the control and treatment groups (Extended Data Fig. 1 and ‘Null results’ section of the Methods).