Advanced analytics uses statistical, machine learning and operations research models for pattern recognition, forecasting, simulation, and optimization. Implementing these methods to generate relevant insights requires new skills and roles in an organization. In addition, these new roles need to be integrated within existing processes to ensure the operationalization of analytics solutions. This post will discuss the roles that are needed for advanced analytics projects and the factors managers should look for when integrating these teams in an organization.
Skills needed
Carrying out advanced analytics projects requires analysts with a strong background in mathematics and statistics. In addition, according to the BARC survey, Advanced & Predictive Analytics, skills in data management for data access and preparation, knowledge of business processes for understanding the task at hand, and the ability to moderate between IT and line of business are the most important skills for conducting advanced analytics projects.
Figure 1: Relevant skills for carrying out advanced analytics projects. Source: BARC Survey Advanced & Predictive Analytics (n=200)
Roles in advanced analytics teams
While initially the data scientist was considered to be the unicorn, covering all these skills, a team approach has become more popular and is more realistic.
Four basic roles are important. The domain expert helps to identify and sharpen the project assignment, provides detailed knowledge of business processes, and helps to identify and to understand relevant data and define how results should be evaluated.
The analytics expert brings in knowledge of statistical methods, machine learning, and optimization. An important part of their work is data preparation. That means that they must be able to transform raw data into technically correct and consistent data. In addition, they need to be able to enhance the value of datasets by feature engineering techniques.
As data preparation can be time-consuming, the data expert has emerged as the third role. They are familiar with database systems, their interfaces and query languages, and know how to source data and design data integration processes, build and maintain data models for analytics, and are able to prepare data so it is ready for analysis.
An often-overlooked role is the data artist. Being responsible for communicating often complex results, this person knows how to tell stories with data and how to use visualizations to enhance the understanding of the business value of advanced analytics. The data artist uses common visualization tools such as business intelligence software, but they also employ other techniques such as infographics.
These four roles are not strictly separated from each other. Often you find data scientists that are also database experts and/or have a good understanding of business processes.
Figure 2: Roles in an advanced analytics team
The organizational integration
Advanced analytics teams are interdisciplinary and often work across different lines of business. An important question is how to integrate these teams within the organization.
Three different approaches are most common. The first option is to integrate analytics teams in IT departments, where they become part of a business iIntelligence competency center (BI CC) or other units such as innovation groups or Big Data CCs. The second option is to integrate the analytics team in the line of business, and the third option is to establish separate data labs.
Each approach has pros and cons with regard to criteria such as flexibility and independency from existing hierarchies and processes, their ability to work across different lines of business, and their ability to operationalize solutions technically and professionally.
Analytics teams in the line of business have the advantage of easy access to domain knowledge. They can carry out adjustments in business processes more easily, which increases acceptance. However, they are limited with regard to their ability to work across different lines of business and their flexibility and independency. Here labs provide a clear advantage, as they are usually independent from existing organizational structures and processes. Experience shows that labs have weaknesses when it comes to model deployment and operationalization.
Handing prototypes to IT has been a problem in many organizations to date. Putting analytics teams right within IT can be an advantage for technical operationalization. The drawbacks are that IT is often not regarded as a driver for innovative solutions. This can hinder the dialogue with the line of business and reduce visibility of analytics teams, making it difficult to incorporate know-how and feedback from the business side. Analytics teams in IT are also not independent from organizational structures. This can reduce flexibility, which is important in this agile and iterative environment.
Figure 3: Different options to integrate advanced analytics teams in an organization
Building resources
In larger or data-intensive companies, it is generally advisable to build internal data science know-how. In the long run, this provides for acceptance of analytics solutions, easier integration in the organization, and sustainable buildup of experience.
Hiring data scientists can be a challenge, but broadening the scope of potential candidates or identifying and training interested employees can solve this problem. External resources can complement internal staff. They can support a quick start of initiatives, when human resources are scarce or not yet available in-house. Working with external support also offers flexibility for phases of high demand and access to market knowledge when it comes to software selection, IT architecture, and best practices.
Once the team is put together and the organizational setup is identified, organizing the process of conducting advanced analytics projects and the interaction with BI teams are important factors to consider. These topics will be covered in the next blog.
Read more from the “Enabling the Data-Driven Enterprise” series