Recommendations 

Just setting up a program can pose the biggest challenge to any organization. Here are a few key considerations and recommendations for improving your program’s quality:

Prevent Culture Shock: Cultural change is never easy in well-established teams. It’s one of the most important human factors to get right when implementing change in a data team. Start by setting everyone’s expectations, including technical players, team managers, and executive stakeholders. Getting everyone committed to the overall goal will help clarify confusion about team roles and responsibilities and build confidence that the process is capable of self-correcting if needed.

Create a Community: Team culture should include a CDSP community for newly established members. It should promote the practice and encourage learning and collaboration while providing ample opportunities for training and sharing thoughts about improving the teams’ skills in the CDSP. The culture should include mentorship by the DS team to transfer best practices as well as on- the-job training for delegated tasks.

Create a Center of Excellence (CoE): Inserting a new team in an already well-established data ecosystem will require planning and coordination for all data team objectives and expectations. Once you understand where and how all the data team members communicate their value, it’s best to include all of them in a main CoE analytical core team.

“A Center of Excellence will be responsible for aligning all teams across the organization on common approaches, architectures, and frameworks. CoE is tasked with creating standards, policies, and procedures. It defines what tools and technologies are used by all teams across the organization.” — Research & Advisory Director at Info-Tech Research Group

Centralization is important for representing the organization’s “single source of truth” about how the CDSP is realized. Some of the CoE’s duties include:

Outlining clearly everyone’s roles and responsibilities.
Defining what tasks are out of scope for all teams.
Working out clear communications among teams.
Deciding what initiatives to work on and the “go/no-go” for all projects.
Determining the size and format of data to be used.
Defining security for those accessing data according to data teams’ requirements and skillsets.

The CoE can and should be the central point for standardizing how all CDSs will work across the organization. It’s a big component, and should not be taken lightly or considered an afterthought.

Define Clear Boundaries: Consider what the CDSs in the LOBs can tolerate with the rest of the team. The newly minted members of your CDSP and your seasoned DSs might not always see eye to eye in the beginning. Your traditional and tenured DSs could resent sharing roles or migrating tasks to CDSs. They might feel encroachment on their “turf” or even feel threats to their job security within the data team.

Don’t Expect Perfection: Expect your newly minted CDSs to make some errors as they get their legs under them and move from walking to running with basic data management and modeling tasks. You will need to implement some safeguards to limit the collateral damage, such as poor data sourcing and cleansing and erroneous model predictions. It’s important to have the team understand this is a new venture and to stress that improvements will be made as part of an iterative and agile process.

Get Early Buy-In: Look to get early buy-in from your data team and technical executives. Executives are busy. You will need to sell them on the benefits of implementing a CDSP, why it’s different from the existing DS team, and how the investment will pay off. For the data team, you should explain why the CDSP needs to happen now and how it will make the analytical flow better with more teams involved. This should also include clear justification for the program and the benefits for all data team members and the broader organization.

Earmark Resources for Keeping Up with the Technology Marketplace: The tools for implementing a CDSP in your organization are widely available and address the full data analytics pipeline. That said, it’s an ever-changing landscape with automated techniques and algorithms in a constant state of development, especially in the open-source community. You should understand your users and the organization’s comfort level for the tools and architecture used in the CDSP implementation.

Start with Existing Tools: Implementing a CDSP means getting more out of what you already have. Don’t feel like you must invest a ton in the beginning. You can look to use some of your existing tools and infrastructure and repurpose them as a sandbox environment for CDS training, experimenting, or even demos.

Find People with the Right Stuff: Those in your organization should share the vision of a well-established and efficiently operated CDSP that works seamlessly within the greater data team. To ensure long-term success, you will need to get such people on board early for some of the following roles:

Senior Sponsorship: This person should be highly motivated to see the company innovate and adapt to the changing market landscape. Preferably, they will come from the executive suite (CDO, CIO) and supply the budget to launch the program.
kljk
Data Champion: This is a big role, filled by a person familiar with the analytical data process. They are committed to making sure everyone does their part in contributing to the program’s success. They are mostly likely on the data team and are passionate about seeing the CDSP come to reality.