The Effects of Enterprise Social Media on Communication Networks

TL;DR

The paper investigates whether Enterprise Social Media Platforms (ESMPs) reshape internal corporate communication beyond traditional channels like email and instant messaging. Using a quasi-experimental difference-in-differences design across 99 adopter firms and a complementary Microsoft case study, it uncovers that ESMP adoption increases overall connectivity by adding many new, weak, one-to-many ties that bridge gaps in the organization. The findings show ESMPs connect employees across distant parts of the hierarchy and distribute influence more democratically, enhancing information flow and cross-unit awareness beyond existing channels. These results suggest ESMPs can counteract remote-work-related silos and foster broader organizational communication while introducing new considerations around tie semantics and forecasting benefits.

Abstract

Enterprise social media platforms (ESMPs) are web-based platforms with standard social media functionality, e.g., communicating with others, posting links and files, liking content, etc., yet all users are part of the same company. The first contribution of this work is the use of a difference-in-differences analysis of $99$ companies to measure the causal impact of ESMPs on companies' communication networks across the full spectrum of communication technologies used within companies: email, instant messaging, and ESMPs. Adoption caused companies' communication networks to grow denser and more well-connected by adding new, novel ties that often, but not exclusively, involve communication from one to many employees. Importantly, some new ties also bridge otherwise separate parts of the corporate communication network. The second contribution of this work, utilizing data on Microsoft's own communication network, is understanding how these communication technologies connect people across the corporate hierarchy. Compared to email and instant messaging, ESMPs excel at connecting nodes distant in the corporate hierarchy both vertically (between leaders and employees) and horizontally (between employees in similar roles but different sectors). Also, influence in ESMPs is more `democratic' than elsewhere, with high-influence nodes well-distributed across the corporate hierarchy. Overall, our results suggest that ESMPs boost information flow within companies and increase employees' attention to what is happening outside their immediate working group, above and beyond email and instant messaging.

Figures (10)

• Figure 1: Study #1: With a difference in differences analysis of 99 companies that suddenly adopted Engage, we study the impact of ESMPs on companies' communication networks across technologies (Instant Messaging, Email, ESMPs). We compare the communication networks of companies that adopted (Company A) or did not adopt (Company B) Engage. Then, we measure the effect of Engage by comparing network properties of the communication network before ($t_1$) and after ($t_2$) some of the companies adopted it. Study #2: Analyzing Microsoft's communication network, we characterize how ESMPs differ from email and instant messaging in bridging the company's hierarchy (right). Modeling Microsoft's corporate hierarchy as a tree, we study how connections in different communication technologies bridge different levels and sectors of the company.
• Figure 2: Engage, the ESMP studied here. Engage offers standard social media functionality (algorithmic feeds, posts, reactions) and is integrated within Microsoft Teams, which offers instant messaging functionalities.
• Figure 3: For a random company in our sample, we show the number of active users across ESM, Email, and Instant Messaging in the 69-week study period considered. A vertical dotted line indicates when the company suddenly adopted Engage, Microsoft's ESMP solution. We use the discontinuity around the adoption of Engage to study the effect of its adoption on the company's communication network (see Sec. \ref{['sec:did']} for methodological details).
• Figure 4: Results of our difference in differences analysis. On the $x$-axis, we show the number of weeks from the intervention, ranging from -5 to 20. Note that interventions are enacted on Week 1 and that Week $-$1 is the reference point. On the $y$-axis, we show the $\log$ effect, i.e., the effect on the outcome (one per panel). The shaded area corresponds to 95% confidence intervals. On the right-hand side of each plot, we also show the aggregated effect ('Agg'), i.e., the average weekly effect across weeks 1 to 20.
• Figure 5: Effects of adopting Engage on metrics associated with tie strength (a-b) and centrality (c-d). Again, on the $x$-axis, we show the number of weeks from the intervention, ranging from -5 to 20; we also show the aggregated effect ('Agg'). Note that outcomes here are not log-transformed -- they should be interpreted as percentage point changes.