Every organization consists of many active formal and informal networks where different types of interactions between people take place. People communicate, share information, ask for help, solve problems, devise plans, brainstorm ideas, and make small talk or lasting friendships – in hundreds and often millions of interactions daily.
When we think of typical knowledge-based organizations, it’s precisely in these interactions that ‘work’ actually happens – even individual contributions are always and inescapably embedded in the collaborative webs of social interactions accomplished with other people.
Over the past decades, research has shown that well-networked organizations – those with unhindered flows of information, knowledge, and mutual trust have better performance, a higher level of organizational learning, better cross-functional cooperation, and higher innovative potential.
From the perspective of the employees, the quality of their networks directly affects their ability to quickly and effectively receive and transmit information, solve problems, manage projects and successfully realize organizational goals – and generally have a more satisfying employee experience.
Despite the unbeatable power of harnessing social networks in organizations, a relatively small number of executive teams think in such terms. This is certainly a gap, but also a huge opportunity for People Teams to reveal an entirely new way to think about managing people in organizations. This article will introduce you to a methodology of capturing relational aspects of organizational life, named Organizational Network Analysis (ONA) and point to some interesting use cases that impact organizational decision-making.
What exactly is ONA?
ONA is a data-driven organizational methodology that analyses the existing patterns of relationships and interactions among employees in organizations. Rob Cross, one of the pioneers in this field and founder of the Connected Commons consortium writes: “ONA can provide an x-ray into the inner workings of an organization – a powerful means to making invisible patterns of information flow and collaboration in strategically important groups visible.”
ONA’s origins go back to the 1930s when sociologists drew their first ‘sociograms’ on paper (graphic representations of the social links a person has) and made calculations of the different metrics that illuminate the behavior of the network and key individuals within them, by hand.
Illustration 1: The first sociogram is attributed to psychiatrist Jacob L Moreno, for his hand-drawn image showing friendship patterns between the boys and the girls in a class of schoolchildren, presented at a medical conference in New York in 1933.
Today, network maps are software-generated, capable of presenting ‘ties’ (relationships) among the ‘nodes’ (people) for even the largest organizations, delivering calculated metrics for both individual and network-level analyses with great speed and precision.
There are roughly speaking two sets of insights that can be revealed through the analyses of organizational networks. One level deals with individual-level positioning concerning the wider group. These network positions – or roles – describe how individual people function within particular social groupings.
For example, individuals with the highest number of social ties (in-degree and out-degree connections) are referred to as central connectors – these are usually employees whose nature of work is to receive and give information to a large number of people. If people are on the edges of networks, they are called isolates – individuals with a very small number of social ties to others.
People receiving large numbers of requests by others to solve complex problems are positioned as knowledge or expert hubs within the organization. If they are positioned as main communicators between two departments, they become brokers, as they function as bridges between different parts of the network.
Pretty quickly, you can see how ONA analysis can identify critical people within the organization who hold important organizational knowledge or control the spread and speed of information sharing, or alternatively, those people who are in some way excluded and disconnected from the organizational flows.
The second group of ONA insights falls into the domain of network structures. Taking a bird’s-eye view of the whole organization, network maps can quickly reveal which parts of the network are isolated from other parts – namely, point to the existence of departmental silos.
It can also inform us of the existence of organizational cliques – small pockets of close-knit people who form communities based on similar interests or attitudes. The overall connectivity within the organization is referred to as network density – the degree to which all actors in the network are interacting with each other and whether those interactions are mutual (degree of reciprocity). Network metrics can also reveal the degree of cross-functional communication among departments, revealing gaps in collaboration and knowledge and information exchange.
ONA data can be combined with many different organizational data sources: gender, performance, compa-ratios, position levels, and tenure – making it possible to reveal both individual-level and structure-level patterns and dynamics in a way that other people analytics approaches are not always equipped to do.
Illustration 2: Different individual-level network roles: bridges, knowledge hubs, and isolates
How do we collect ONA data?
In our decade-long experience with running ONAs, this is the most frequent stumbling block for the in-house People Teams. But collecting ONA data can be a rather straightforward and painless process. Before clarifying how it’s done, let’s talk about some initial decisions you need to make. There are two data collection methods in ONA: passive and active.
Passive data collection involves generating data without the active participation of the employees. This includes metadata – or logs – collected from phone calls / conversations over the phone , email, or text activity. This data already exists in the organization but obtaining access to it is often difficult and ethically questionable (if done without the explicit consent of the employees).
Second, there are important issues with the quality of passive data: while being able to capture all digital communication, the logs are not capable of classifying the content or the context of exchanges. For instance, we don’t know if two employees are exchanging messages to solve a problem, brainstorm an idea, complain about a procedure, or chit-chat about the weather. This lack of content/context makes it difficult to generate insightful recommendations regarding the meaning of relationships observed.
Active data collection involves using direct input from employees through a series of questions in a survey. These questions allow us to target our inquiries to specific flows – of expertise, trust, problem-solving, energizing exchanges, innovation, and ideas and other chosen domains. Think of these questions as a process of ‘nomination’ – each employee in the organization lists or selects people with whom they usually exchange some form of communication. Examples of questions include:
- Who do you most often contact to solve a difficult work-related problem? (expertise flows)
- Which people do you usually go to when you want to discuss a new idea? (innovation flows)
- Which people do you usually contact to obtain information necessary for your everyday work? (information flows)
- Exchanges with these people in my organization leave me energized (energizing relationships)
and many other questions, depending on the specific domains of interest.
Each ‘pair’ in this crowdsourced process represents a connection (‘edge’) in the resulting network map. The number of participating employees will equal the number of ‘nodes’ or actors in the map, with each node receiving a certain number of ‘in-degree’ (incoming) and ‘out-degree’ (outgoing) connections. Mathematically speaking, there are n*(n-1) possible connections in every (directional) network. Quick calculation for an organization size of 100 people yields an astonishing 9900 possible connections.
Illustration 3: Calculating the number of possible connections within a network
These in-degree and out-degree connections are the foundation for further analyses – this is why it’s critical to the validity of ONA analysis that the participation rate of employees remains high (95-99%, depending on the size of the organization). Considering that anonymous surveys have at best 75% response rates, securing high participation can be a tall order for any People Team.
Before attempting to run ONA, People Teams should establish a high level of ‘data trust’ – employee confidence that their (personal) data will never be used to their detriment, but rather as a powerful tool to improve ways of working together.
One advantage of ONA surveys is that they are relatively painless for the participants – a well-designed survey will take your employee no more than 10-15 minutes of attention to complete. See below for David Green’s nifty overview of active vs. passive ONA differences.
Use Case for Organizational Network Analysis
Use cases for ONA are practically limitless, especially when the basic ONA data is cross-analyzed with other sources of organizational data. Here we’ll address a recent case:
Whether a start-up is in hyper-growth mode or a more stable business, organizations often encounter serious problems with departmental silos and a lack of adequate cross-functional collaboration and alignment. These risks have only intensified during the pandemic. While companies increasingly turn to remote-first work policies (significantly boosting employee morale!), remote work can, unfortunately, act as a destabilizing factor in the creation of well-functioning organizational networks.
In our case, during the pandemic, we’ve enjoyed tremendous growth, and the team nearly doubled in size. All the new joiners were onboarded in a fully remote setup at 19+ locations worldwide. Making connections and building networks, once an organic process, especially for our co-located teams, now had to become very intentional because the teams were growing so fast and were dispersed all around the world. We knew we were facing challenges in maintaining cross-functional alignment but were not sure how to best target our resources to mitigate these growing pains.
We applied the Organizational Network Analysis in 3 domains we identified as the most important in the context of hyper scaling. We conducted:
Communication/Information flow network analysis to tell us…
a.) What does the overall organizational connectivity look like, after nearly 2 years of the pandemic and remote work? Which departments have low levels of information sharing with others, i.e., are siloed?
a.) Which people are key connectors and/or bottlenecks in the organizational information flow?
b.) Which employees needed more help to integrate into the organizational network and make meaningful connections?
Expertise network analysis to tell us…
a.) Which employees represent central knowledge hubs and where are the retention threats in the company?
b.) Where are the organization’s hidden talents? (highly sought-after experts, who were not necessarily sufficiently recognized to reflect their importance within the company)
Trust network to tell us…
h.) What is the overall level of interpersonal trust in the organization?
i.) Where are critical points in the organization where trust needed to be improved?
Data collection efforts took us 3 weeks – 199 employees from 24 global teams filled out the survey (95% of our employee base). Because of the strong focus on cross-functional collaboration, our survey also included one qualitative open-ended question, asking employees for their suggestions on how to improve this aspect. We were very careful to explain the fundamentals of the analysis openly and transparently – why we are conducting it, how employee answers will be used in it, and how it will improve the way we work together.
All the data was collected using a survey platform (alchemer.com in our case), exported into an Excel file, and transformed to fit the format required by the visualization software (kumu.io in our case). Many other analysis software options are available on the market: Panalyt, Polinode, and Trustsphere, to name just a few.
Some of the key ONA takeaways and actions
ONA provided us with many powerful insights into the internal workings of our organization.
Network level insights:
#1 Our overall communication network density was 13%, with reciprocity of 40% and an average network size of 26. Our trust network density was 7%, with reciprocity of 23% and an average network size of 12. Our expertise network was 6%, with reciprocity of 15% and an average links-per-node of 14.
These collaboration patterns revealed high inter-departmental silo effects with linkages and trusted relationships mostly residing within, not across different teams. Average network sizes were relatively small, clustering according to departmental and to a smaller degree, geographical (co-location) lines.
Illustration 4. Communication Network Map, color-coded by departments, node size scaled by in-degree connections
#2 On a network structure level, there were very few strong communication ties between the business and technology (blue) parts of the organization. To some degree, this is considered a usual finding in the tech start-up world. What our maps and metrics showed us, however, is that those linkages were dependent on only several, disproportionately burdened individuals (bridges – dark blue circles).
This presents a risk for the network functioning for several reasons: one, the bridges with this much collaborative overload can easily burn out and become flight risks; second, they slow down communication and decision-making processes as they can easily bottleneck; third, their departure would temporarily fracture the network until new or back-up linkages were formed, threatening the stability and efficiency of the entire network.
Illustration 5. Two clusters of the communication map indicate siloed departments
#3 We then took a closer look at the collaboration patterns between each of our departments by calculating the communication density between every two departments. The resulting matrix showed us where the biggest cross-functional collaboration gaps were, but also which teams were overburdened with information requests from the rest of the organization.
Illustration 6. A simplified cross-functional collaboration matrix that shows how many connections (out of 100% possible) are realized between every two departments
Given these results, our initial focus was on increasing overall cross-functional connectivity and stability of the network. We’ve created a #mittoconnections program that doubled down on both the formal and informal relationship-building within the company. This included:
a.) focused in-person strategy alignment sessions for departments lowest on collaboration density
b.) internal mentorship program (“Power of Two”) to connect experts with mentees, allowing for a much more dynamic flow of expertise and knowledge within the organization,
c.) creation of project groups/task forces requiring cross-functional representation from affected departments,
d.) random social buddy initiative to allow people to create social ties with ‘distant’ parts of the organization,
e.) activation of the ‘social team budgets’ for informal gatherings
f.) an internal requirement for teams to organize company-wide Lunch & Learns after every conference and/or training.
Individual-level ONA insights
Looking at the individual-level metrics, insights get much more specific.
#1 Identification of key people (information, knowledge, and trust hubs) across the company
Looking at the in-degree and out-degree metrics, ONA allows individual ranking of the highest sought-after and the most connected individuals in the organization, for each type of map generated. Due to the size of their networks and their positioning, these ‘hubs’ and ‘connectors’ exhibit a high degree of influence in the organization, holding on to critical knowledge and controlling the speed and spread of information.
This part of network analysis always yields surprises, allowing executive teams not just to validate some prior assumptions and/or assessments of top performers, but also to spot hidden talents – people whose organizational importance is not always formally recognized through pay and hierarchy. Cross-analysing ONA data with the compa-ratios and performance data allowed us to intervene at these critical spots and prevent knowledge loss. It also allowed us to reduce the collaborative overload of affected individuals by activating their ‘deputies’ (people who hold the next-in-line network status) in the organizational knowledge-sharing processes.
Individuals who are high on both expertise and trust often represent key influencers in the network (thought to be both credible and likable). These employees are natural change ambassadors, and it’s wise to include them in various cross-functional projects where the need for obtaining buy-in is high, ensuring projects’ acceptance and success across the departmental lines. The high connectors in the network also impact the “emotional climate” of the company – due to their strategic position, they can influence the opinions of many people they come in contact with on a daily basis.
# 2 Identification of isolates (people on the network periphery)
We used the low in-degree and out-degree metrics to identify peripheral people – employees who are for some reason at the very edges of networks relatively disconnected from the rest of their colleagues. Cross-analyzing low connectivity with tenure and performance data, we came to some interesting conclusions. One, peripheral nodes are often juniors or new joiners who are still being onboarded and therefore not yet well integrated within the information flows. One important action point that emerges from ONA analysis is the importance of creating those initial patterns of connectivity as a part of people’s onboarding. Introduction of new hires to key people early on – through both formal and informal means – increases their chances of success by placing them further towards the center of information and knowledge flows.
The second finding was also interesting and actionable: when looking at people with longer tenure and low connectivity, the top 5 isolates with lower performance scores had an 80% chance of leaving the company 6 months after the analysis. This illustrates the predictive power of ONA to spot flight and performance risks just by virtue of employees’ social embeddedness in organizational networks (discussed at more length in organizational literature elsewhere).
#3 Identification of ‘bottlenecks’ and individuals with collaborative overload risk
Taking into account all network requests (communication, expertise, innovation, etc.), ONA can very quickly and with a high degree of reliability identify which employees are at risk of burn-out due to collaborative overload.
We cross-referenced the communication and the expertise maps and found there was a strong correlation between them, with expertise reciprocity ranking at a low 15%. Those individuals who sought out more information were at the same time considered to be the main knowledge hubs in the organization. This correlation showed us that we had many single owners of critical knowledge and that they were often overburdened with a high volume of incoming information requests, which was increasing their risk for burnout as well as creating bottlenecks in the decision-making processes.
After the ONA analysis
Managing “what you do with the data” is as important as how you collect and analyze it. Throughout the process, we implemented ‘data checks’ with some of the trusted stakeholders in the organization. These data checks allowed us to brainstorm some of the initial findings and come to the correct interpretations. For example, even though some cross-departmental collaboration may appear low, if those departments generally do not need to collaborate to a high degree, this finding was considered normal and not in need of improvement.
At different stages of the process, we delivered ONA insights to departmental leaders to inform their strategies and decision-making. We plan to repeat the analysis on an annual level, tracking the changes in the internal metrics and obtaining further buy-in for the action plans and initiatives that resulted from the research. Often, ONA’s results are quite different than what leaders would expect based on formal organizational charts and defined business processes, and our analysis added value to processes such as talent acquisition, talent development, succession strategies, retention, engagement and overall improvement of efficiency and alignment.
Maja Ninković Shapera, VP People, Mitto
Nada Krstić, Senior Manager, People & Culture, Mitto
Publication: HR World Magazine No. 8 (2022)