Navigating the new normal: Examining coattendance in a hybrid work environment

Abstract

The recent rise of hybrid work poses novel challenges for synchronizing in-office work schedules. Using anonymized building access data, we quantified coattendance patterns among ~43k employees at a large global technology company. We used two-way fixed effects regression models to investigate the association between an employee’s presence in the office and that of their manager and teammates. Our analysis shows that employee in-person attendance was 29% higher when their manager was present. Moreover, a 1-SD increase in the share of teammates who were present yielded a 16% increase in the individual employee’s attendance. We also observed greater coattendance among employees who were recently hired, have a Corporate or Operations role, or work in shared office spaces. Thus, we find evidence of some voluntary alignment of work schedules. Companies could bolster such organic coordination by leveraging digital scheduling tools or providing guidance specifically aimed at increasing coattendance.

As the COVID-19 pandemic subsides, much of the workforce—28% of US full-time employees as of March 2023 (1)—has transitioned to hybrid work arrangements, wherein employees work from home on some days and in the office on others. Hybrid work seeks to combine the benefits of in-office and remote work. Face-to-face encounters, resulting from in-office work, can improve team collaboration (2) and connection (3) and facilitate activities like brainstorming (4). Remote work, on the other hand, offers flexibility, saved commute time, and enhanced individual focus time (5–7); it also enables access to a broader talent pool (8) and potential cost savings for organizations and employees (6). Hybrid work aims to retain some of the remote work advantages, while capturing some of the collaboration and productivity benefits of in-person time (9). However, hybrid work poses new challenges, particularly the synchronization of in-office workdays among employees (10). Misaligned work schedules can lead to missed opportunities for in-person interactions (11), which is one of the top reasons employees come to the office (12). By examining attendance patterns in hybrid settings, organizations can learn from scheduling preferences and assess coordination levels. Such learnings, in turn, can help organizations design better hybrid work policies to foster teamwork and coattendance.

In this study, we quantify attendance patterns among ~43k employees at a global technology company. We leverage anonymized building access data, obtained from electronic badges used by employees at three major office sites worldwide (US headquarters, India, and Ireland), along with data on each employee’s job category, workspace type, new-hire status, and anonymized manager’s identifier. Throughout our analyses, we ensured the data were fully anonymized, and no individual information was revealed. We focus on the tendency of an employee to coattend with their direct manager and teammates (i.e., people reporting to the same manager). Although our approach does not allow us to determine who is influencing whom to go to the office due to the reflexive nature of coattendance (13), it enables us to reliably assess alignment of in-office work schedules. To uncover factors that affect coattendance patterns, we investigate whether such correlations differ by employee job category, new-hire status, and workspace type. Our overall objective is to understand the extent and heterogeneity of team coordination in office attendance among the employees we study and knowledge workers more broadly.

Results

After fully closing in March 2020, the company’s offices partially reopened in 2021 at all three considered sites. On February 28, 2022, offices at the US headquarters fully reopened, marking the end of our 60-d “pre-period” and the start of a 30-d transition period intended for employees to adapt their schedules; our 60-d “post-period” follows. After reopening, employees were encouraged to go onsite more, but generally no attendance requirements were enforced.

We began our analysis by estimating conditional average attendance nonparametrically, assessing how the likelihood of an employee working in the office varies with their manager’s and teammates’ presence on that day. Smoothed attendance patterns are shown under four distinct scenarios in Fig. 1, reflecting whether an employee’s manager or at least one teammate was onsite. Employees were more likely to be in the office when their manager or teammates were present and even more likely when both were there. Although attendance increased at the start of the transition period, the four lines in Fig. 1 shifted in tandem, suggesting that there continued to be partial coordination in attendance.

Fig. 1.

Employees’ office attendance rates in 2022, given the presence of their manager or teammates. The conditional average attendance for employees at the US headquarters was estimated nonparametrically and smoothed using a 5-d rolling window. Shaded bands represent 95% CIs with team-level clustering.

Next, we employed a two-way fixed effects (TWFE) regression model to investigate the association between an employee’s presence in the office and that of their manager and teammates. The TWFE model uses individual fixed effects to control for each employee’s baseline attendance rate and date fixed effects to control for day-to-day fluctuations in company-wide attendance, due to factors such as free lunch or events happening on that day. The two variables of interest in the model are the manager’s attendance (binary variable) and the share of teammates who are in the office (continuous variable, unlike Fig. 1, which uses a binary variable for whether at least 1 teammate attends). Since attendance of the full team was very rare, even during the post-period, we report the change associated with a 1-SD increase (~0.27) in teammates’ attendance instead of the coefficient itself.

Results of TWFE regression models for the US headquarters, run separately for the pre- and post-period, are shown in Fig. 2A. During the post-period, an employee was 29% (7.7 percentage points or pp) more likely to be present onsite when their manager was and 16% (4.2pp) more likely when the share of their teammates present onsite increased by one SD. During the pre-period, the association with manager’s attendance was proportionally larger than in the post-period, at 42% (P < 0.01); however, the baseline attendance rate was also much lower (9pp pre vs. 27pp post). In contrast, the proportional association with teammates’ attendance in the pre-period (13%) was qualitatively similar to that of the post-period, albeit statistically significantly smaller (P < 0.01). These results suggest that coordination between employees and their managers decreased from the pre- to the post-period, whereas coordination with teammates remained stable. A possible explanation is that during the pre-period, some employees only went to the office on days when they had 1:1 meetings with their managers, but in the post-period, they also went on other days, uncoordinated with their manager.

Fig. 2.

Relative association with the manager’s and teammates’ attendance for models run separately by (A) period, (B) location, (C) job category, (D) new-hire status, and (E) workspace type. All graphs except (B) are for employees at the US headquarters only. Eng, PM, and Corp refer to Engineering & Research, Product & Program Management, and Corporate & Operations, respectively. We report TWFE coefficients of teammates’ attendance multiplied by the corresponding SD 0.17 (Pre), 0.27 (Post), 0.27 (US), 0.24 (India), 0.26 (Ireland), 0.26 (Eng), 0.28 (PM), 0.29 (Corp), 0.26 (New), 0.27 (Existing), 0.26 (Closed), 0.26 (Neighborhood), divided by baseline attendance. Error bars represent 95% CIs with team-level clustering.

We performed the same analysis for employees located in India and Ireland, using country-specific reopening dates (Fig. 2B). In the post-period, we observed similar associations between a given employee’s presence onsite and that of their manager across all three countries (26% in India and 29% in Ireland and the United States). The association with teammates’ attendance was somewhat larger in India (21%) and in Ireland (24%) than in the United States (16%).^* We discuss possible factors contributing to heterogeneity in peer effects by location, including differing workplace cultures and distributions of employees across job categories, in SI Appendix. The existence of meaningful coattendance patterns at multiple locations worldwide supports the robustness of our findings, emphasizing the importance of the manager’s and the team’s presence in driving the employee’s attendance across geographical and sociocultural settings.

Depending on their job category, employees may have different reasons for office attendance. For instance, nontechnical staff may prioritize seamless operations and direct communication facilitated by in-person interactions, while technical personnel might value onsite meetings for improved knowledge transfer and problem-solving. Focusing on the US headquarters in the post-period, we analyzed these associations for the top three categories by number of employees: Engineering & Research (Eng), Product & Program Management (PM), and Corporate & Operations (Corp). Both in relative and absolute terms, the association with manager’s attendance for Corp (44%, 10.3pp) and PM (38%, 8.9pp) were notably larger than for Eng (24%, 6.9pp) (Fig. 2C). The relative increases associated with a 1-SD increase in teammates’ attendance were more similar across job categories, but the ordering was identical (Corp 19%, PM 16%, Eng 15%; SDs were similar across categories).^† This finding suggests that employees in Corp and PM may rely more on managerial guidance and face-to-face communication for decision-making, whereas employees in Eng might work more independently.

Next, we categorized employees into “new hires” vs “existing hires,” based on their start date at the company (after vs. before the pandemic began in March 2020). New hires may prefer frequent office visits to acclimate themselves to company culture and seek out mentorship, whereas experienced employees might opt for a flexible work location, reserving office visits for critical meetings or collaborative projects. Focusing on Eng employees to avoid confounding by job category, we found substantial differences between these two subgroups, as shown in Fig. 2D. The association with manager’s attendance was significantly larger for new versus existing hires (29% vs. 23%, on a relative scale, P < 0.01). A similar difference was observed for teammates’ attendance, with a relative association of 18% among new hires and 14% among existing hires (P < 0.01). We conjecture that new hires may either experience more peer or manager pressure to be onsite when their colleagues are, or actively seek in-office time to learn from colleagues, form better ties, and improve their social capital (14, 15).

Finally, we examined differences in coattendance patterns by workspace type. Some employees work from open, neighborhood spaces shared with their teammates, whereas others have closed, private offices. We hypothesized that the two groups had different reasons to go to the office: Employees with a closed office might be more motivated by a quiet space conducive to focused work and confidential discussions, while those working in a neighborhood space might be more motivated by social interactions with colleagues. For estimation, we again focused on Eng employees, who are more evenly distributed across the two major workspace types. While employees assigned to neighborhood spaces had a 5.0pp lower average attendance rate than employees with closed offices, their attendance was more correlated with that of their manager (neighborhood: 29%, 7.6pp; closed office: 20%, 6.6pp) and teammates (neighborhood: 17%, 4.5pp; closed office: 13%, 4.0pp), see Fig. 2E. Employees working in neighborhood spaces might benefit more from the collaborative environment, resulting in greater motivation to coattend, or might experience greater peer pressure to go onsite when others do.

Discussion

We interpret our findings as evidence of “organic coordination” among employees. Even with the ongoing COVID-19 pandemic and limited attendance expectations, employees went to the office more on the days when their manager and teammates were present. In practice, coordination may have taken multiple forms. For example, some teams agreed on a day of the week to prioritize for in-person meetings among team members; other teams fostered the habit of posting to a chat channel on Friday their planned in-office workdays for the following week. There was evidence of this coordination even before the site’s full reopening, as employees’ attendance was already correlated with that of their managers and teammates during the pre-period. In sum, even without any dedicated digital scheduling tool, or official encouragement, employees managed to partially align their in-office workdays.

We studied the time period when offices first reopened at this company, in the aftermath of the Omicron wave. The pervasiveness of COVID-19 concerns may explain the relatively low attendance rates. Since then, office attendance rates at this company and nationally have increased. More recently, many companies have announced new policies to increase office attendance. Our research suggests that companies implementing hybrid work policies like the one examined here, or asking people to come in any 3 days per week, may wish to complement such approaches with efforts to increase bottom-up coordination; these could take the form of team-level discussions and explicit coordination of regular in-office days, software tools to facilitate coordination among coworkers, or both. Companies requiring office attendance on certain days of the week (e.g., Tuesday, Wednesday, and Thursday) are missing many of the flexibility and office-use efficiency benefits of hybrid work but could still benefit from bottom-up coordination for employees to align their schedules when they need to do in-office work on nonrequired days. Our heterogeneous findings, across job categories, new-hire status, and workspace types, suggest tailored strategies to encourage coordination could also be beneficial.

Materials and Methods

Descriptions of our data and regression models are in SI Appendix.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

R and Python scripts used to estimate statistical models and generate figures, regression coefficients including absolute and relative effects and standard errors data have been deposited in Github (16). Data cannot be shared due to employee privacy and other legal restrictions, raw confidential data underlying this study is not available for public sharing.