Organizing analytics capabilities within an organization
The term Big data is defined as “data whose volume, velocity, and variety make it difficult for an organization to manage, analyze, and extract value using current methods”, whereas analytics consists of the “processes of sense-making and useful insights generation”. It has increased its popularity threefold between the years of 2012-2015 and has been a very hot digital trend for the past 6 years. There is little debate about how big data & analytics can support the strategic goals of the organization and create value. However, how best to organize your analytics capabilities in the organization, is still a question companies are struggling to answer.
Following the issue, the interdisciplinary figure of “data scientist” has emerged combining skills from business, programming, data visualization & communication, machine learning and analytics abilities. In 2012, Harvard Business Review called it “The Sexiest Job of the 21st century” and soon enough it became companies’ main human resource for big data analytics initiatives. The diffusion of data scientists in Italy has been on a high increase as many companies were adopting new organizational models to manage them.
New Organizational Models
Observatory for Big Data Analytics of Politecnico di Milano investigated through an international survey submitted to data scientists and developed the following 4 organizational models to manage these resources:
- Centralized: Under this model, there are data science teams independently structured who serve to the entire organization and report to the Chief Data Scientist (CDS) that decides and prioritizes the projects to work on. This model is applicable for organizations that operate with limited data scientist resources to appoint for each business line in the long-term.
- Business Driven: This model suggests fully-embedding of data scientists under the business units such as marketing, operations, finance etc. Moreover, there is no specific role as the CDS to coordinate the individual data scientists.
- Matrix: The Matrix model embeds data scientists under business lines similar to Business Driven model but require them also to report to the CDS instead of only business unit leaders. Overall, a medium level of data science capabilities works best with this model in order to be effective.
- Hybrid: Being the most complex model, it combines the independent organization structure of the Centralized model and the embedding of data scientists under business lines as in the Matrix model. This is most applicable for large companies who would like to manage different type of tasks utilizing both of the two different structures at the same time. Expectedly, there can be potential coordination difficulties among structures.
Survey On Department Positioning
Politecnico di Milano surveyed data scientists about the department they are located in and came up with this pie chart:
Source: International Survey of Data Scientists, Observatory for Big Data Analytics of Politecnico di Milano, 2016
The chart suggests that 30% of the data scientists are located within a business line, which can be interpreted that their company is adopting the Business Driven or the Matrix model. Similar in size, 26% of data scientists work under an independently structured team specific for data science activities as in the Centralized Model, which could as well be the Hybrid Model. Finally, 15% of the interviewees are located under the IT department and 8% of them are consultants hired from external providers. These findings illustrate that no specific model is widely adopted, thus can be observed in companies. Rather, it tells us that:
Companies are still discovering the best organizational ways to manage their data scientist resources specific to their business needs, capabilities and size.
Some are investing more, increasing their analytics capabilities and some still need to get these competences externally.
To conclude, it is very important for companies to identify the relevant big data & analytics processes within the organization. This will in turn help them decide on the best possible governance structure to extract the maximum business value from big data and enhance competitive advantage.