The impact of COVID-19 on digital communication patterns

The COVID-19 global pandemic disrupted the way organizations function, just as it disrupted life more generally. As infections increased, governments shut down physical work sites and required shelter-in-place orders, prompting a rapid shift to working from home for many knowledge workers. This abrupt transition occurred amid large-scale economic and social upheaval, at a time when organizational coordination, decision-making, and productivity were highly consequential. This paper provides a large-scale analysis of how formal digital communication patterns changed in the early stages of the pandemic, addressing a gap in systematic evidence on day-to-day work activity changes. The study explores how the pandemic altered interaction patterns—measured through meeting and email activity metrics—as organizations rapidly moved to remote work. The analysis uses de-identified meta-data from a provider of digital communication solutions, covering 3,143,270 users across 21,478 de-identified firms in 16 large metropolitan areas. The meta-data include frequency of emails and meetings, meeting size and duration, number of email recipients, timing of emails, and related dimensions. The data’s geographic and longitudinal precision enables tracking communication before and throughout the first stage of the pandemic. Sixteen metropolitan statistical areas (MSAs) with government-mandated lockdowns provide clear breakpoints indicating shifts to remote work, with lockdowns between March 8 (Milan) and March 25 (Washington, DC), and observation windows spanning 8 weeks before to 8 weeks after lockdown to study behavioral change.

The paper synthesizes adjacent literature to infer potential changes in digital communication during forced remote work. Prior work on remote work largely examined voluntary, smaller-scale transitions in more stable contexts, limiting generalizability to the pandemic. Evidence on cross-firm digital communication is scarce, often focusing on a single firm or self-reports rather than large-scale meta-data. Research also offers limited guidance on how meeting dimensions (frequency, size, length) interrelate, though they likely co-vary and reflect organizational trade-offs between coordination and individual work time.

• Meeting frequency: Digital channels are less information-dense than face-to-face interactions, suggesting newly virtual teams might meet more and email more to compensate for lost social/contextual cues and to enforce accountability in virtual settings. However, virtual meetings can be more cognitively demanding and less effective, with technological hurdles potentially dampening any unconditional increase in meeting frequency as teams modulate to avoid diminishing returns.
• Meeting size: Norms for meeting size differ across organizations and affect effectiveness and inclusiveness. Remote work could increase meeting size (no room-capacity constraints; broader relevance of issues) or decrease it (more one-on-ones for accountability; attentional limits and distraction risks in larger virtual meetings).
• Meeting length: Evidence is mixed. Virtual fatigue and attention limits could shorten meetings, especially if more check-ins are used. Conversely, pandemic-era problems might require longer meetings, and online inefficiencies might necessitate more time.
• Email activity: Email can complement or substitute for meetings. Asynchronous, one-to-many communication may make some tasks more efficient via email, while others still require meetings. The extent to which email substitutes for or complements meetings under remote work is an open question. Given the lack of definitive theory and the novelty of forced, widespread remote transitions, the authors refrain from formal hypotheses and proceed descriptively.

Data come from a large IT services provider licensing digital communication tools globally. The provider supplied aggregated, de-identified meeting and email meta-data for 3,143,270 users across 21,478 firms in 16 MSAs with government-mandated lockdowns. Measures include: email frequency (counts of internal and external emails; average count of recipients) and meeting activity (average count of meetings, average meeting duration, average attendees per meeting). Broader work-pattern measures include average workday span (time from first to last communication in a day), cumulative meeting hours per person per day, and emails sent outside regular business hours. Cleaning steps: dropped meetings with a single attendee or over 8 hours (to exclude out-of-office/personal calendar blocks), removed meetings with >250 attendees (to exclude company-wide notices/spam), and limited emails to internal messages between same-domain employees. MSAs included each city’s central business district and surrounding suburbs/townships >100,000 population. Inclusion criteria for the 16 cities: at least 50,000 active users across 500 firms; clear work-from-home orders; orders enacted between March 8–28, 2020, to align timing (excluding Asian cities with earlier lockdowns). Variables were pre-aggregated by the provider at the MSA-day level; no user-level or personally identifiable data were accessed by researchers. A secondary dataset provided email aggregates by industry (SIC-1) and organization size (small <250; medium 250–500; large 1000–2500; enterprise 2500+) for robustness; industry-level meeting data were unavailable. Empirical strategy: An OLS event-study framework compares post- versus pre-lockdown periods and examines weekly dynamics. Models include MSA fixed effects (to control for time-invariant MSA differences) and day-of-week indicators; standard errors clustered at the MSA level. The “lockdown week” is a 7-day window centered on the lockdown date; the base week is one week prior to the lockdown week (anticipation effects noted). No evidence of pre-trends was found prior to the base week. The study window spans 8 weeks pre- and 8 weeks post-lockdown; additional analyses extend email measures up to 9 months post-lockdown. Robustness checks include weighting by user counts, weekly vs daily aggregations, alternative reference week (8 weeks pre-lockdown), MSA- and industry-level heterogeneity, and checks on user base stability.

Meeting activity: Compared to pre-pandemic levels, total meetings per person increased by +12.9% (95% CI: +11.4% to +14.4%), about +0.8 meetings per person per day; average meeting duration decreased by -20.1% (95% CI: -23.0% to -17.1%), roughly -12.1 minutes per meeting; average attendees per meeting increased by +13.5% (95% CI: +10.6% to +16.5%), about +2.1 people per meeting. The net effect reduced total hours spent in meetings by -11.5% (95% CI: -14.3% to -8.7%), approximately -18.6 minutes per person per day. Weekly event-study estimates show consistent increases in meeting count/size and decreases in duration for each week after lockdown, yielding sustained reductions in weekly meeting hours relative to the base week. Email activity: Internal emails sent increased by +5.2% (95% CI: +3.0% to +7.6%), about +1.4 emails per person per day. The average number of recipients per email increased by +2.9% (95% CI: +0.3% to +5.5%), about +0.25 recipients per email. External emails showed no significant change. Extending up to nine months post-lockdown, internal email volume remained significantly above pre-lockdown levels, whereas the average number of recipients returned to baseline by about the third month. Weekly dynamics indicate internal emails spike in the lockdown week and decline over subsequent weeks, returning to pre-lockdown levels around week four. Work patterns: The average workday span increased by +48.5 minutes (+8.2%, 95% CI: +7.1% to +9.3%). Emails sent after business hours increased by +8.3% (95% CI: +4.0% to +12.7%), about +0.63 emails per person per day. Elevated workday span persisted through the eight post-lockdown weeks. Robustness and heterogeneity: Results (except recipients) are robust to weighting by user counts and to weekly aggregation. Patterns are broadly similar across industries and sizes; Europe appears more negatively impacted when controlling for holidays, possibly reflecting stricter lockdowns or pre-existing work-life balance norms. Services (excluding financial services) showed slower recovery in email intensity, possibly due to reduced demand for in-person services. User base composition remained stable.

Findings suggest organizations rebalanced communication modalities under remote work, increasing the frequency and inclusiveness of formal communications (more meetings, more attendees, more internal emails/recipients) while decreasing the time intensity of meetings (shorter durations) and, overall, time spent in meetings. This pattern is consistent with leveraging virtual communication as an imperfect substitute for lost face-to-face interactions, potentially freeing time for focused work while still coordinating at scale. Internal email increased—and persisted—while external email did not, possibly reflecting a turn inward to synchronize internal processes and accountability, and/or the presence of automated external communications that were unaffected by remote status. Increases in meeting size/frequency may have aimed to reduce social isolation, maintain identification, and disseminate information broadly in the absence of informal in-office interactions, while shorter meeting lengths align with attention limits and home-based distractions in virtual settings. Despite reduced meeting hours, a greater number of meetings and attendees raises coordination costs and interruptions, potentially affecting well-being. Spillovers beyond standard hours are evident: employees’ workday spans lengthened and after-hours emails increased. Possible mechanisms include flexibility stigma prompting longer hours to signal commitment, blurred work–home boundaries, and more irregular schedules as workers accommodate home demands. Organizations appear proactive: communication patterns began shifting at least a week before formal lockdowns. The study showcases the value of passively collected communications meta-data for timely organizational insights and highlights managerial trade-offs in setting meeting size, length, and frequency under remote or hybrid arrangements.

COVID-19 lockdowns were associated with increases in the total volume of meeting and email activity, larger meetings, shorter meeting durations, decreased total time in meetings, and longer workday spans. By analyzing digital communication across many firms and regions, the study extends research that uses communication meta-data to understand organizational behavior and contributes to the virtual work literature by examining a mandatory, organization-wide transition rather than voluntary, small-scale shifts. The results underscore that organizations can rapidly recalibrate communication along multiple dimensions in response to shocks and that some changes, particularly increased internal email volume, can persist months later. Future research should investigate optimal balances among meeting frequency, size, and length across contexts; explore cross-platform substitutions beyond email and calendared meetings; disentangle the roles of remote work, macroeconomic shocks, and non-work behaviors; and assess how different remote, hybrid, or mixed-mode arrangements shape communication networks and outcomes.

• Scope of data: Only formal communications captured by the provider’s email and calendar platform are included. Informal meetings, instant messaging, phone calls, and other platforms (e.g., Slack, Microsoft Teams) are excluded, so cross-platform substitutions are not observed.
• Concurrent shocks: Lockdowns coincide with multiple factors—forced remote work, macroeconomic demand shocks, and non-work behavior changes. The study cannot disentangle their individual effects; estimates reflect the joint impact of co-occurring forces. Heterogeneity in lockdown strictness and organizational responses (e.g., hybrid vs mixed-mode) is not fully captured.
• Measurement limitations: Calendar-based meeting data may misclassify some events (despite cleaning); scheduled durations may not match actual durations. Such errors likely vary across firms but are not expected to bias results systematically.
• Industry-level meeting data unavailable: Industry heterogeneity analyses for meetings could not be conducted due to data constraints (only email by industry/size available).