Analytics capabilities facilitate companies to gather, process and distribute information from inside and outside the firm more broadly. Not only does this capability support day to day decision-making, it can potentially play an important role in innovation – the process by which companies invent the building blocks that will someday lead to new products or services.
While it is easy to associate innovation with the lone inventor painstakingly discovering a brilliant idea [such as Thomas Edison literally making the lightbulb go (and stay) on], modern innovation is a collaborative process involving large numbers of people bringing their unique expertise to solve new problems. In fact, even Edison had his assistants and collaborators that were integral to his success. Moreover, while it is exciting to think of innovation as developing ‘breakthrough’ technologies, most innovation represents a series of small steps where existing ideas are combined in new ways, often across disciplines. It is in these processes – facilitating the sharing of information and gathering existing knowledge to synthesize new knowledge where analytics has the potential to accelerate the innovative process.
Not all organizations structure their innovation process in the same way. In an insightful article, Mark Wilson at FastCompany observed that two of the most innovative modern companies have fundamentally different ways of organizing their innovators. Using a data visualization tool for looking at named inventors on patents, he showed that Google had a highly decentralized structure for their innovators with widely distributed sets of collaborative relationships. By contrast, many of Apple’s patents are created by small closely-knit groups of collaborators. It turns out this contrast extends well beyond Google and Apple – some firms have a highly centralized innovative structure (like Apple) where others are more decentralized (like Google). For instance, two highly innovative firms, Sanofi and Roche show a similar, observable difference in innovation:
This structure matters for at least two reasons. First, different structures have different demands for information and different advantages and disadvantages in obtaining information. Centralized structures can share information readily across the development team and dedicate resources to gather specific external information that can be applicable to the whole firm as opposed to stakeholders in a single group. By contrast, decentralized organizations face greater costs of external information gathering and coordination and tend to focus on information with immediate benefit to the group. Analytics technology can lower these costs, providing an advantage to decentralized firms.
Second, these two types of structures tend to produce different types of innovations with centralized structures more likely to create breakthrough, novel innovations and decentralized structures more likely to develop innovations reusing or combining existing knowledge in new ways. Again, analytics will tend to be especially valuable for decentralized firms since recombination is naturally more data intensive (after all, there is no data to analyze about products ideas that do not yet exist). Thus, we hypothesized that while analytics can potentially support the innovative process generally, it was likely to be more useful in decentralized firms and especially helpful in the production of innovations involving reuse or recombination of existing technologies.
Using data on patents from the USPTO we were able to measure both the innovative structure as well as the types of innovations produced by more than 1,800 large firms. It turns out that machine learning community detection algorithms are very effective in distinguishing centralized versus decentralized innovation (the Apples versus the Googles) and allows us to translate the relationships shown on the network graphs to quantitative measure of decentralization. We combined this data with measures of data analytics capabilities (captured as the number of employees with data skills) derived from public and private sources of the skills of the employees that work for these firms. We estimate a variety of statistical models attempting to identify correlations as well as causal relationships between innovation structure, analytics and innovative output.
Overall, consistent with our story, firms that have a more decentralized innovative structure hire more employees with data analytics skills than firms with a centralized structure. This is perhaps because they also receive greater benefits from their analytics investments – firms that are in the top third of both decentralization and analytics investments are about 2-3% more productive than their counterparts that do not combine analytics and decentralization. Furthermore, we find that firms that combine analytics and decentralization are more innovative, but only in areas where it is likely where data provides an advantage – the identification of new combinations of existing technologies or the develop of novel technologies that exist in the broader marketplace but have not previously been developed by the company. This is consistent with analytics providing an advantage in gathering, processing and bringing in existing knowledge from outside the company and finding new ways to combine existing ideas to make new ideas.
Ultimately, the bottom line of this analysis is that analytics can help improve the bottom line and can do so by helping firms get more out of their innovative efforts. But it is not just a matter of having analytics capability – these capabilities are most effective when combined with other organizational practices that can leverage data.
To read more, see our working paper at SSRN (“Data Analytics Supports Decentralized Innovation”) which is forthcoming at Management Science.