Every business leader has heard the advice: to innovate, build partnerships. Tap into outside expertise. Join the ecosystem. And the numbers back this up. According to recent surveys cited in the research, 43% of business leaders say external partners are a key driver of innovation, and 94% of tech executives consider innovation partnerships essential to their strategy.
But what if your partners all know each other? And what if their partners know each other too? A new investigation published in the Journal of Business Research suggests that the shape of the broader network surrounding a firm can make or break its ability to launch new products. And more connections are not always better.
The question hiding in plain sight
Most research on business networks has focused on a firm’s immediate partners, which researchers call the “ego-network.” But real business relationships extend further. Your partner has partners. Those partners have partners too. When all those connections are densely woven together, information flows freely. That sounds good, until you realize your proprietary knowledge can flow just as freely in the other direction.
Eric Schaap, who conducted this research as part of his doctoral work at Maastricht University in the Netherlands, set out with his co-authors to examine this extended web of relationships. The team wanted to know how “higher-order network density,” meaning the degree to which a firm’s partners and their partners are all interconnected, shapes a firm’s likelihood of launching a new product.
They also wanted to understand how this interacts with something called “cognitive distance.” In plain terms, cognitive distance is the difference in knowledge, focus, and expertise between two firms. A software company and a textile manufacturer have high cognitive distance. Two biotech startups working on similar gene therapies have low cognitive distance.
A new way to map business relationships
Traditional studies of business networks rely on patent data or alliance databases, both of which have limitations. Patents tend to favor certain industries. Alliance databases often miss informal relationships. Surveys can undercount actual network size.
The research team tried something different. They combined data from the Community Innovation Survey, a standardized questionnaire used across EU member states to track whether firms have launched new products, with information scraped from the websites of Dutch companies with at least 10 employees.
The logic is simple but creative. When one company’s website links to another company’s website, that hyperlink is often a signal of a real business relationship. About 18% of the connections the team found were reciprocated, meaning both websites linked to each other, suggesting deliberate partnership signaling rather than casual mentions.
Using an open-source tool called ARGUS, the team scraped up to 25 pages from each company’s website. They then mapped the connections between firms, ending up with a network of 2,320 companies connected by 356 hyperlinks. An algorithm called Infomap was used to identify “communities,” which are naturally occurring clusters of interconnected firms within the larger network.
For measuring cognitive distance, the researchers analyzed the text on each company’s website. Using a machine-learning technique called Doc2Vec, they converted website text into numerical representations that capture the meaning and context of the words. Then they measured how similar or different each company’s website was compared to others in the same community.
What the data showed
After running the analysis on 304 firms that belonged to detected communities, the research revealed a curved relationship between network density and the likelihood of introducing a new product. Innovation likelihood climbed as density increased, up to a point. After that point, more density was associated with lower innovation likelihood.
To put numbers on it, the team found that moving from a sparsely connected community to a moderately connected one was associated with roughly a seven-percentage-point bump in the probability of a firm launching a new product. But pushing further into very high density territory appeared to reverse those gains.
The story got more interesting when the team layered in cognitive distance. When firms in a community had knowledge bases very different from one another (high cognitive distance), the curve became steeper. The benefits of moderate density were amplified, and so were the costs of excessive density. When firms were cognitively similar, the relationship was much flatter. Density simply mattered less.
The researchers interpret this through a framework developed by scholar Dovev Lavie, which describes three types of value, or “rents,” that firms can capture from networks: value from their own resources, value from direct partnerships, and value that spills over from the broader network. At moderate density, firms can balance all three. At low density, they lack the trust and information flows needed. At high density, their proprietary knowledge leaks out faster than they can capture value from it.
What this means for managers and policymakers
For business leaders, the practical takeaway is that evaluating partnerships requires looking beyond immediate relationships. A new partner might seem attractive on its own, but if that partner sits inside a tightly woven cluster of companies with knowledge bases very different from yours, joining that network could mean your innovations leak out to competitors before you can capitalize on them.
The methodology itself offers something useful. Companies can regularly map their network position by analyzing hyperlinks and website content, without needing cooperation from partners or access to confidential data. This gives managers a way to monitor how their network is evolving.
For policymakers, the message complicates the common assumption that denser innovation clusters are always better. Governments often promote network formation to drive regional innovation. But at very high density levels, firms may struggle to retain exclusive advantages from their research, as happened in Denmark’s life sciences cluster where knowledge spread evenly across members.
Some caveats are worth noting. The network map was built from a snapshot in 2021, which creates a slight mismatch with the 2020 survey data. The cognitive distance analysis only included Dutch-language websites, potentially excluding insights from English-language content. And the measure of innovation is binary, capturing whether a firm launched a new product but not how significant it was.
Still, the findings add nuance to a question that matters for nearly every business pursuing growth through partnership. More connections can help, but only up to a point, and only depending on who those connections are.
