Picture a small technology firm that joins a bustling regional business cluster hoping to spark its next product. The pitch is appealing: more partners, more shared knowledge, more chances to combine ideas into something new. But there is a catch that many managers overlook. It is not just about who a company knows. It is about how tightly everyone else in that web is connected to each other, including partners the firm has never dealt with directly.
A study published in the Journal of Business Research examined this question and found that when it comes to the density of these extended networks, more is not always better. The researchers report an inverted U-shaped pattern: a moderate level of interconnectedness is linked to the highest likelihood of launching a new product, while both sparse and overly dense networks are associated with lower odds.
The blind spot in network research
Most earlier work on business networks and innovation has focused on a firm’s immediate circle, what researchers call the ego-network. That means looking only at a company’s direct partners and how connected those direct partners are to one another. Eric Schaap, who completed the work at Maastricht University in the Netherlands, and his colleagues argue that this view misses a large part of the picture.
Their focus is on what they call higher-order network density. In plain terms, this measures how interconnected a firm’s broader web of partners and partners-of-partners is, while deliberately setting aside the firm’s own direct ties. The reasoning is that a company can control who it links to directly, but it has little say over how densely the surrounding players are wired together. That surrounding structure shapes how easily knowledge flows, and how easily it leaks.
The authors describe a tension at the heart of dense networks. On one hand, tight interconnection builds trust, cooperation, and shared understanding, which can help a firm absorb complex knowledge. On the other hand, in a very dense network, as the authors put it by quoting earlier work, “everyone knows what everyone knows.” Proprietary insights spread quickly to everyone, including competitors, which can erode the advantage a firm gains from access to new ideas.
A second ingredient: how different partners are
The second concept the study leans on is cognitive distance, which refers to how dissimilar firms are in their knowledge bases, organizational focus, and technical expertise. A gene-therapy startup and a traditional device maker sit far apart cognitively; two payment-processing firms using similar technology sit close together.
Cognitive distance matters because it determines whether the knowledge flowing through a network is genuinely novel or largely redundant. The researchers propose that cognitive distance amplifies both the upside and the downside of network density. When partners are very different from one another, the benefits of trust-building in a moderately dense network grow, but so do the risks that valuable proprietary knowledge will leak in a network that is too dense.
Mapping a network from company websites
To test these ideas, the team needed a way to map connections across many industries at once, something traditional patent or survey data struggle to do. Their solution was to scrape the websites of Dutch companies with at least ten employees.
The approach rests on two assumptions supported by prior research. First, a hyperlink from one firm’s website to another’s tends to signal a collaborative relationship, so the team used these links to draw a map of who is connected to whom. Second, the text on a company’s website reflects its knowledge and worldview, so the team analyzed that text to gauge cognitive distance. To measure how similar two firms’ website texts were, they used a technique called Doc2Vec, which converts documents into strings of numbers so that texts with similar meaning end up mathematically close together.
They then combined this web data with the Community Innovation Survey, a standardized questionnaire run across the European Union that asks firms whether they introduced a new or improved product in the previous three years. This survey served as the measure of innovation outcomes.
From an initial network of 2,320 firms linked by 356 hyperlinks, the researchers used an algorithm to detect naturally occurring clusters, which they call communities. This let them study higher-order structures without having to analyze the entire national network. After filtering, they identified 57 communities containing 304 firms, which formed the sample for their main statistical analysis, a logistic regression predicting whether a firm launched a product innovation.
What the numbers showed
The analysis offered evidence for the inverted U-shape. Higher-order network density was positively associated with the likelihood of product innovation up to a point, after which the relationship turned negative. As a rough gauge of size, the researchers estimate that each one-unit increase in density corresponded to roughly a 28-percentage-point change in the likelihood of innovation, though density in their sample varies within a much narrower range.
The picture became richer once cognitive distance entered the model. At high levels of cognitive distance, when surrounding firms held very different knowledge, the inverted U-shape was pronounced: the benefits of moderate density were strong, but so were the penalties of excessive density, and the peak of the curve shifted toward lower density levels. At low cognitive distance, when firms shared similar knowledge, the relationship was relatively flat, meaning density mattered much less.
The researchers interpret this through a framework of “rents,” a term for the value a firm captures from its resources and relationships. They argue that moderate density lets a firm balance the value from its own capabilities, its direct partnerships, and the wider network, while both very low and very high density throw that balance off.
What managers might take from this
The authors suggest that firms should look beyond their direct partners when deciding how to position themselves. Their web-scraping method offers one practical tool: because it relies on publicly available website data, a manager could map a firm’s extended network and estimate cognitive distances repeatedly over time, without needing cooperation from partners or access to confidential alliance records.
The takeaway they emphasize is contextual. When surrounding firms are cognitively distant, network structure deserves close attention, because the effects of density are magnified in both directions. When surrounding firms are similar, companies have more freedom in how they connect. For policymakers who often promote dense clusters to drive regional growth, the authors caution that excessive interconnectedness can spread knowledge so widely that individual firms struggle to hold onto an advantage.
Reasons for caution
Several limits are worth keeping in mind. The innovation measure is a simple yes-or-no from a survey, which may not capture the intensity of research in industries with long development cycles. The cognitive distance analysis relied on Dutch-language webpages, and the team notes that firms with English content appeared marginally more likely to be innovative, a gap that could shift results. The main analysis rested on 304 firms, and several individual coefficients were not statistically significant on their own.
Perhaps most important, the network and the survey were captured at slightly different times, and the design is largely observational. The authors ran a check using archived versions of websites and found little sign that innovation drove the formation of new links, but they still describe the relationships as associations rather than proven cause and effect. Establishing causation, they note, would require tracking these networks over time.
