The Chief Data Officer (CDO) is not a new position – Capital One reportedly had a CDO all the way back in 2002. But only recently has it become a mainstream, business-critical role for enterprises. In a recent study by NewVantage Partners, 65% of surveyed companies have a CDO position within their organization. And as data and AI continue to shape nearly every industry, the role of the CDO is shifting with it.
In this blog, we dive into the evolution of the CDO and, as data-driven leadership becomes the key to business success, the prescriptive steps required for building data and AI-centric organizations.
Evolution of responsibilities of the CDO
At a high level, the CDO is responsible for envisioning and executing the data strategy across all business functions. Over time, the CDO role has evolved with global economic and regulatory changes and continuous advancements in technology. According to Gartner, with the uptick in data and analytics (D&A) use cases, we have entered into the fourth stage of the CDO evolution, defined as:
- Stage 1: Primarily focus on data management strategies and best practices. This can be either centralized or decentralized under distinct business units.
- Stage 2: Utilize analytics along with data management, which helps with “the offense” since the CDO can now design holistic data management strategies that best enable the use of data analytics.
- Stage 3: Drive digital transformation or business transformation.
- Stage 4: Manage profit and loss (P&L) using D&A instead of merely focusing on creating the D&A products.
As CDOs across organizations transition to “stage four,” their approach and metrics of success ultimately come down to one question: are they optimizing for only a defensive or an offensive data strategy that is built on sound data management practices? Let’s walk through what each of these scenarios means.
In a defensive approach, one focuses on ETL, data management, infrastructure, regulatory and compliance issues. These are the foundations of a good data platform since they are concerned with protecting the data and controlling access to the data.
An offensive strategy, on the other hand, focuses on the products and uses D&A to drive decision-making for P&L. CDOs are integral to decision-making and work with the business units. These are revenue-generating strategies or strategies concerned with increasing customer satisfaction. In this scenario, it’s important to note that data management is still critical; the difference is that CDOs are now looking beyond pure data management in a defensive strategy to build on those practices and create value out of data.
According to Harvard Business Review, offensive strategies tend to be more real-time since they focus on sales and marketing, while defensive strategies focus on compliance and the legal aspects of data management. Ideally, one would want to balance the competing features of control (defensive) and flexibility (offensive) of data to utilize the data effectively. While it is easy to lock up the data in a silo, this is counterproductive since now no value can be derived from the data to bring in new customers and increase revenue. It is important to point out that a good defensive strategy is essential before one can transition to offense, (i.e. one needs good data management and governance policies in place before one can start to put the data to use with machine learning models and analytics).
Technology enabling the transition from defense to offense
According to a recent MIT Tech Review study, “Building a High-Performance Data and AI Organization,” which sought to understand the characteristics of high-performing data-driven organizations, most high-performing organizations are focused on implementing ML solutions with their data, whereas lower-performing organizations still struggle with implementing data management strategies. High-performing organizations prioritized the following in their quest to become a more data-driven organization:
- Reduced data duplication
- Fast and easy access to data
- Improved data quality
- Minimize the hurdles in cross-functional collaboration
- Perform analytics ‘in-place’
Data management – the single source of truth – can evolve to derive versions of this, but the provenance has to be clear. First of all, data governance has to be strictly enforced to make sure that this is done consistently. Data management can be either centralized or decentralized under each individual business unit. The advantage of a centralized data management policy is stronger controls over your data, however, a decentralized approach results in more flexible access to data, making it easier to apply analytics and generate insights from the data. Stronger controls or data governance can result in reduced data duplication and better data quality while flexibility helps with faster and easier access to data.
A new data architecture is emerging, the lakehouse architecture, which brings the best of data warehouses and data lakes into a single unified platform for all data, analytics and AI. A lakehouse architecture both helps CDOs effectively check the box on data management and build am offensive data and AI strategy. Modern lakehouse architectures build on existing open lakes to seamlessly add comprehensive data management, supporting analytics and ML across all business units using all enterprise data. With lakehouse architectures, CDOs can reduce risk by enabling fine-grained access controls for data governance, functionality typically not possible with data lakes. Data can quickly and accurately be updated in the data lake to comply with regulations like GDPR and maintain better data governance through audit logging – even in multi-cloud environments. CDOs now shift focus on the exciting value add initiatives of creating compelling business insights and turning data into products.
The pace of investment in D&A products is increasing.
There is greater adoption of ML and deep learning solutions. However, there are two primary issues that plague a data science team. One is avoiding the inevitable staleness associated with moving data from the source to the model building platform. Ideally, a framework that allows models to be trained on data ‘in-place’ or where the data resides is needed. The second issue relates to handing off ML models from the data scientists to the engineering teams (i.e. there has to be a seamless way to productionize ML models). In order to minimize these issues, there must either be a close collaboration between the data scientists and the engineering team or a seamless process for the model-building team to deploy it to production without involving the engineering team.
Through this shift, it’s important to note that the importance of governance never goes away but rather the shift is additive, with a solid offensive strategy executed on a strong foundation of governance.
Challenges for the modern CDO
“Culture eats strategy for breakfast” by Peter Drucker. According to the NewVantage Partners executive survey on Big Data and AI, the same can be said for almost 92% of organizations with respect to being data-driven. Organizations, while aspiring to be data-driven, can be hesitant to make the necessary changes required to becoming a data-driven organization. CDOs need to be mindful of this while trying to work across functional units. As reported in the NewVantage survey, more executives believe that the CDO role should belong to someone who is an ‘insider’ as opposed to an external agent of change. This is possibly signaling the desire for someone who is in sync with the company culture, thereby minimizing the hurdles associated with bringing about change in the data culture.
To achieve this goal, a CDO needs the adoption of its products across the entire company, not just individual teams. This requires close cooperation with and endorsement from data-driven leaders in the C-suite, otherwise, these initiatives are bound to fail. According to data leader veteran Sol Rashidi of Estee Lauder, the best approach for increasing adoption is to start with a “prototype” and get buy-in from the business leaders before proceeding. This creates tangible alignment with the business interests as opposed to abstract goals that are difficult to align on. The key here, according to Rashidi, is to de-emphasize the details and focus on business outcomes and the value derived from D&A products rather than just the technical capabilities.
The path to success
The most critical step for CDOs to enable data and AI at scale is to develop a comprehensive strategy with buy-in from stakeholders across the organization. This strategy focuses on achieving maximum success by leveraging people, processes, data and technology to ultimately drive measurable business results against your corporate priorities. The strategy serves as a set of principles that every member of your organization can reference when making business decisions. It should cover the roles and responsibilities of teams within the organization for capturing, storing, curating and processing data — including the resources (labor and budget).
To help guide CDOs and other data leaders looking to transition to an offensive, business-focused data strategy, we’ve compiled a list of 10 key considerations. You can see the full list and details in the guide “Enable Data and AI to Transform your Organization”:
- What are the overall goals, timeline and appetite for the initiative?
- How do you identify, evaluate and prioritize use cases that actually provide a significant ROI?
- How do you create high-performing teams and empower your business analyst, data scientist, machine learning and data engineering talent?
- How can you future-proof your technology investment with a modern cloud-based data architecture?
- How can you satisfy the GDPR, the CCCPA, and other emerging data compliance and governance regulations?
- How do you guarantee data quality and enable secure data access and sharing of all your data across the organization?
- How do you streamline the user experience (UX), improve collaboration, and simplify the complexity of your tooling?
- How do you make informed build vs. buy decisions and ensure you are focusing your limited resources on the most important problems?
- How do you establish the initial budgets, allocate and optimize costs based on SLAs and usage patterns?
- What are the best practices for moving into production, and how do you measure progress, rate of adoption, and user satisfaction?
A strategy should clearly answer these 10 questions and more, and it should be captured in a living document, owned and governed by the CDO, and made available for everyone in the organization to review and provide feedback. The strategy will evolve based on the changing business and/or technology landscape — but it should serve as the North Star for how you will navigate the many decisions and trade-offs that you will need to make over the course of the transformation.
To learn more about how CDO’s and data leaders are evolving their roles and their views on the future of their data strategies check out the Champions of Data + AI series and the new report from MIT Tech Review Insights: Building a High-Performance Data and AI Organization.