Like most complex analysis, social network analysis is iterative. The steps outlined below are meant to serve as a high-level guide to the process rather than a strict sequencing.

Key steps in conducting a social network analysis include:

1. Define your learning question
2. Define network parameters
3. Engage the network
4. Collect data
5. Analyze findings
6. Share results

What do you want to better understand?

One of the first steps when conducting a network analysis is to clearly define your learning question(s). What specifically do you want to better understand about the relationships among organizations in a particular network? Learning questions that may help you better understand a local system include:

• Who are the local organizations that other actors "go to" for help and assistance?
• How do organizations interact and collaborate with one another around a common goal?
  • Goals of networks may vary widely, from the very general to the granular.
  • For example, a general network goal might be to “help people to obtain jobs,” or “make migrant populations more resilient.” Examples of a narrower goal might be to, “reduce the incidence of HIV among unemployed males aged 18-25 in Timbuktu.” This goal, no matter how general or specific, should be in what all of the organizations in the network are working towards.

As there is typically no opportunity for a re-do once data has been collected, it is important to think carefully about both respondent (node) and relationship (edge) attributes to be collected in order to answer your learning questions. This could include:

• functional groupings of actors (e.g. association, government, union, etc.)
• demographics (# of employees, women-led, etc.)
• subnetworks (e.g. industry)

Another critical step in the SNA process is conducting stakeholder consultations to define key network parameters. The parameters are used to refine and contextualize the survey instrument and the data collection process, and are fundamental to ensuring that the network analysis will respond to the learning question(s). Some key parameters to establish are:

• Network Boundary: Which actors should be included in the network when collecting data? A clear definition of the network boundary must be established to capture data on as many network actors as possible, without including non-members. This typically includes clarification of a common goal of all actors, a geographic boundary, and potentially other characteristics dependent on the network.
• Actor Attributes: What other actor characteristics are important to network analysis? Actor attributes allow us to segment data and analyze subgroups of network actors based on those characteristics. This typically includes type of organization or institution, technical area interests, and size or age of the organization.
• Relationship Content: What types of relationships should be evaluated? Based on the learning objectives, types may include information sharing, resource sharing, collaboration, client-supplier, advice-seeking, or others. The quality of the relationships can also be evaluated by collecting data such as frequency of communication, level of the organization at which the relationship exists (e.g., executive, administrative, operational), utility, strength, or trust.
• Establishing a Relevant Timeframe – Just as boundaries must be set on whom to include in the study, also time boundaries must be set on which links to include and which to exclude between those in a network. For example, should all links between network members over the last 5 years, the last 3 years, the last year, or the last 6-months be included?
• Target Respondents: Who should be interviewed / surveyed within each network actor? Even for SNAs, evaluating relationships among organizational or institutional actors, relationships are managed by individuals. The analysis is most accurate when the correct individuals (e.g., executive director, board members, program directors, operational managers) respond to the survey instrument.

Before finalizing the survey instrument and beginning the data collection process, it is important to meet with representatives from organizations that will likely be a part of the SNA to discuss the study. This, will allow for ensuring a clear understanding of the purpose of the study, the type of insights that will come from it, and how to complete the questionnaire (or process for conducting the in-person interviews). Engaging the network can be done in a workshop setting with representatives from all the organizations expected to meet the boundary for the SNA, small group discussions, or online webinars.

Primary Data

• A survey instrument needs to be designed and piloted to collect data on actors and their relationships. The questionnaires can be completed online using a computer or smart phone, or through phone or in-person interviews.
• While it may be preferable to be able to pre-identify all actors in the network prior to data collection, this is often not possible and important actors can be missed. In these cases, as known network actors are surveyed, additional network members are identified either through a "snowball" approach (the network expands until all network actors are identified) or an ego-alter approach (network expands a set number of times). Follow-up phone calls are generally required to get completed questionnaires or schedule an interview time.
• Survey results will need to be reviewed and cleaned in preparation for the data analysis phase. In particular, it is critically important that an organization's name appears in the exact same manner throughout the data.

Secondary Data

• Depending on the parameters defined, network analysis can utilize secondary data from organizational or public records such as contracts and agreements, emails and other communications, or meeting attendance sheets.
• Additional secondary data helps contextualize the results of the analysis. While SNA results are useful on their own, they also complement results of other tools to provide a deeper understanding of the individual actor and system-level constraints and opportunities.

Data collection and analysis can be completed once for a snapshot of the structural opportunities and constraints in a network, or can be repeated at several points in time to evaluate network evolution.

Once survey data collection is complete, network analysis software can be used to help examine the network as a whole (macro-level), and individual organizations (ego-level). It is typically most effective to first analyze the macro-level network for a few key metrics, which subsequently guides the analysis of individual organizations themselves. On this basis of this, results are analyzed and scores assigned to various indicators being tracked by a project or by network members themselves.

Several key variables are typically analyzed at the network level and for specific actors:

• Density: measures the number of ties between actors indicating the level of connectedness within the network. The density of a network may give us insights into such phenomena as the speed at which information diffuses among the nodes, and the extent to which actors have high levels of social capital and/or social constraint. It is measured by dividing the number of existing connections with the total number of all possible connections. If values have been assigned to these ties (e.g. strength, closeness), then the total sum of those actual values is divided by the total possible number in the network.
• Centrality: indicates which actors are most engaged and which are peripheral
• Reciprocity: measures the extent to which relationships reported by one actor are confirmed by the other actor. A network that has a predominance of null or reciprocated ties over asymmetric connections may be a more "equal" or "stable" network than one with a predominance of asymmetric connections (which might be more of a hierarchy).
• Distance: calculates the average number of steps for any network actor to reach another actor
• Clusters: indicate the existence of sub-groups of actors that are completely interconnected (and often only loosely connected to the rest of the network, if at all). Where distances are great, it may take a long time for information to diffuse across a population. Those actors who are closer to more others may be able to exert more power than those who are more distant.

Longitudinal data can be used to analyze changes in the network over time. As a network evolves, the ONA is able to track impact on local systems against activities undertaken by projects. As a result, network analysis feedback loops enable program implementers to appropriately calibrate their interventions as they progress and learn from them.

Below are some best practices for sharing results:

• Share-back with those that participated in the analysis itself to help to facilitate collective action processes. (In many cases this leads to a request to conduct more in-depth analysis on their own organization / networks.)
• Whenever possible, publicly post research and results, including any survey instrument utilized.
  • Other options include, sharing results via workshops, blog posts, or other write-ups.
• If possibly, conduct analysis using UCINET, NodeXL, or other open source platforms to make data sharing easier.
• ALWAYS ensure that any sharing you do complies with IRB requirements!