Under conditions of demand uncertainty coupled with stable supply, human asset specificity, task complexity, and frequency of exchange, organizational fields develop structural embeddedness. In contrast to relational embeddedness, which essentially refers to the quality and depth of a single dyadic tie, structural embeddedness can be defined as the extent to which a "dyad’s mutual contacts are connected to one another" (Granovetter, 1992: 35). This means that organizations do not just have relationships with each other but also with the same third parties; thus, many parties are linked indirectly by third parties. Structural embeddedness is a function of how many participants interact with one another, how likely future interactions are among participants, and how likely participants are to talk about these interactions (Granovetter, 1985, 1992). Due to decoupling, subcontractors and professionals move frequently among firms and fellow professionals in networks; this links differing groups together and spreads information about third parties among those within the network. This allows information, norms, and common understandings to move across group boundaries (Friedkin, 1982; Granovetter, 1973, 1982). In addition, since parties’ mutual contacts know or know of one another, they have a greater interest in the information and are more likely to share it with one another. The more structural embeddedness there is in a network, the more information about each player is known to all the other players and the more constraints there are on each player’s behavior (Mayhew, 1968; Burt 1992).

Structural embeddedness is critical for understanding how social mechanisms coordinate and safeguard exchanges in networks. Structural embeddedness is a conduit for diffusing values and norms which enhance coordination among autonomous units, and for diffusing information about parties behaviors and strategies which enhances safeguarding customized exchanges. Thus, structural embeddedness allows parties to use implicit and open-ended contracts for customized, complex exchanges under conditions of demand uncertainty. Structural embeddedness enables social mechanisms such as restricted access, macroculture, collective sanctions, and reputation to coordinate and safeguard exchanges. It makes restricted access possible because it provides information so that parties know with whom to exchange and whom to avoid. Negative gossip by third parties about a party’s uncooperative behavior significantly reduces the likelihood of direct relations whereas positive gossip strengthens the likelihood of direct relations (Burt & Knez, 1995). Gulati’s (1995) work on alliances also shows that parties gather information regarding potential opportunities, synergies, and exchange partners through indirect links identifiable through structural embeddedness. Since structural embeddedness diffuses information throughout a system, it also facilitates the development of macroculture, which is common values, norms, and beliefs shared across firms, because parties share perceptions and understandings (Pfeffer & Leblebici, 1973) and institutionalize these beliefs, norms, and values through this interaction (DiMaggio & Powell, 1983). Collective sanctions are not possible without structural embeddedness since parties must know about misfeasance in order to act jointly to condemn or ostracize perpetrators. For reputation to be effective information about parties’ behavior must flow throughout the system. Thus,

Proposition 2: Structural embeddedness provides the basis for social mechanisms to adapt, coordinate, and safeguard exchanges; thus, its presence enhances the likelihood of network governance emerging and thriving in rapidly changing markets for complex, customized tasks.

Too much embeddedness may create its own set of problems. Uzzi (1996b) suggests that over-embeddedness in relational embeddedness (i.e., many strong ties and few weak ties) can lead to feuding, choking off of novel information from other parts of the industry, and welfare-like support of weak network members. Essentially, over-reliance on strong ties tends to tight, relatively isolated cliques which are not well integrated with the rest of the industry (Granovetter 1973). The optimal level of structural embeddedness in terms of overall fitness of the network may be in an intermediate range where parties are neither too tightly connected to fragment social connections nor too loosely connected that they are unaware of who has information to provide, or who knows whom in order to request the information. The optimal level of structural embeddedness may be determined by network size and is an important empirical question.

In contrast, if one focuses strictly on dyadic characteristics of transactions and ignores structure, one assumes that exchanges among dyads can be understood in isolation "... without reference to the nature of other ties in the network or how they fit together" (Wellman 1991:35-36). In fact, drawing conclusions about network governance from strictly dyadic data amounts to falling for the well-known ecological fallacy (Rousseau, 1985) where data from one level of analysis (e.g., correlation of smoking and cancer rates across nations) are interpreted as if they were drawn from another level of analysis (e.g., effect of smoking on cancer in individuals).