– Venkatesh Ravirala
One recent HBR article articulated “Why Chief Data and AI Officers are Set Up to Fail” [1] and provided good recommendations on “How to Fix it.” I would also advocate that when ROI and successful business growth outcomes are of paramount importance, corporations should lead with the paradigm of “Decisions-Optimization Focus” and not just “Data-Driven” culture enthusiasm. Data-driven culture transformation arguably benefits from “Make data everyone’s business.” In contrast, optimal decisions are better achieved by empowering and deferring decision-choices analyses and recommendation to seasoned analytics experts.
Business strategy and corporate investments decision-making needs to be a shared responsibility between business administrators (holding CXO titles) as well as analytics leaders. Especially, executives including Chief Analytics Officers (CAOs) need a demonstrable track record built upon delivering good decisions through hands-on data analyses, solutions design, AI/ML predictions, automation delivery, experimentation rigor, and related business decisions-optimization.
Over the last couple of decades, a confluence of rapid advancements in big data and specialized cloud compute have brought AI/ML to the forefront of business transformation. Authoritative leaders in the domain of data-technology and business administration have evangelized use and monetization of data extensively. Consequently, not just the data but actual analytic-solutions design has become increasingly complex. CAOs tackling challenging business problems through innovation and automation now require deep technical expertise and manage increasingly complex challenges and CXO demands across enterprise. In many cases though, despite significant efforts and investment, many corporations still find themselves caught behind the eight ball lacking data integration [2], immature data governance [3], and poor data quality [4]. Contrary to expectations of valuable data being a high leverage to achieve growth and business optimization, corporate leaders often complain that data immaturity is their biggest stumbling block.
Over the past 12-months, CEOs across many industry sectors are in cost-control and even more drastic massive layoffs mode, questioning the return on investments both in staff and data technology [4]. With the emergence of ChatGPT, the topics of AI and GPT and related euphoria of automation as well as fear of hallucinations have consumed all the oxygen at dozens of Data, Analytics, and AI conferences. Especially over the last four years, it has often been stated that the vast majority (some sources quote even 85%) of IT and Analytics projects have not delivered positive or sufficient ROI to business.
How to Organize Data and Analytics Functions?
Stepping back a bit, during the 1990s, the first decade of my career, traditional information technology C‑level roles of CIO, CTO, CPO and CISO have played a crucial role enabling data-centric functions. They ensured that “Data-Driven” culture can help improve business outcomes by creating and consuming analytic-insights. As opposed to IT, it is primarily the other CXO leadership across the other traditional business functions (such as commerce, consumer, marketing, supply-chain, medical, digital, diversity) that drove business strategy and decision-making. They employed small expert teams to use advanced predictive, optimization and simulation techniques to build specific solutions, be it demand forecasting, price optimization, or promotional marketing investment related customer analytics.
As my career progressed towards holding executive leadership and advisory consultant roles, I have directly reported to and delivered to the mandate of CIO, CTO, CFO, CMO, CCO, CPO, and CEO. There never seemed to be a good permanent reporting structure for a CAO role.
More than a dozen times I have built Advanced Analytics teams ground-up as a “Center-of-Excellence” supporting niche demands depending on which CXOs sponsored investments. Teams remained relatively small, ranging from a dozen for mid-sized companies to three hundred at Fortune 100. Primary responsibilities were Innovation, Research, Solutions Development, Measurement, Deployment, and End-to-End Business Value Delivery. Sometimes there were specific top and bottom-line growth and monetary targets to meet, and at other times the staff I managed were part of an enablement cost-center. Other supportive analytics functions for functional business units (as opposed to enterprise-wide) such as data delivery, business intelligence gathering and reporting, and interactive and exploratory visualization dashboards for insights were secondary responsibility. This demanded close partnership and collaboration with numerous IT and OT teams managing broader organization enablement and business functions at enterprise scale.
As the demand for data science exploded over the last decade, organizations are again leaning towards IT-centralized organizational design. In this design, thousands of data engineering, product, analytics, program management, operations and related resources work within a single IT C-level organization. The “Analytics” leadership is again shifting back to be owned by IT C-level roles.
These hub-and-spoke teams have many pros and cons, and there is no one-model-fits-all recommendation. Every organization has to take many different parameters and leadership dynamics into consideration when choosing to lean towards one. Those interested in learning more about six organizing models for analytics teams and read the guidance from IIA [6]. There is however one specific guidance I offer after considering a crucial question of Who Owns Data Delivery?
Who Owns Data Delivery?
Understandably, ‘data’ has been called the ‘new oil’, and it is the primary responsibility of information technology (IT) in conjunction operational technology (OT) leaders to deliver data not in original crude form across silos but sufficiently aggregated, cleansed, labelled, schema-designed (where appropriate), validated, secured, and made cost-effectively consumable using friendlier tools. Arguably, IT is yet to accomplish this mission. When it comes to data-delivery, business value is to be measured not in terms of business growth and increased profits but whether IT/OT is sufficiently low-cost and enabled ease-of-use. The question ought to be: what would the non-IT/OT CXOs be willing to pay for that data-as-a-service if all of IT/OT were an independent subsidiary? Some companies that have entirely out-sourced IT functions have a better handle on this investment decision.
The value that business can derive further utilizing that data-as-a-service depends on many more strategies and value-add solutions. In other words, value of data as oil is to be measured akin to market price of the oil supply, and not get caught up in creating many expectations through use cases and ROI analyses about how that oil can launch a rocket-ship with satellites. This measure will significantly improve the success rate of IT data delivery.
Additional business use cases value delivery ought to be the charter of Chief Analytics Officers (CAO) who must succeed in partnering with other CXOs. With AI taking center-stage, this role has now been extended to encompass AI-related responsibilities. For greater success, organizing ownership of training data enablement by CDO, operations and infrastructure technology management by CIO/CTO, and analytic value delivery by CAO is warranted.
Is the CDAIO Role Sustainable?
Mr Bean and Ms Sagraves [1] suggested that for a CDAIO role “Rather than technology and infrastructure problems, the focus of the role should have been on business outcomes.” As I explained above, the role of IT/OT, including a CDO ought to focus on successful and cost-effective delivery of data-as-a-service. For most non-tech businesses, IT CXOs focusing on business outcomes through analytics solutions would be a stretch. This would be analogous to expecting a quarterback in American football to repeatedly rush towards the end-zone as opposed to focusing on throwing forward passes.
In my 25-years of experience across seven industries, and in consulting roles helping over one-hundred companies, I rarely ever came across executive leaders who are competent to deliver on Data, Business Analytics, AI/ML Analytics, and Business Value by owning all functions. The focus and competencies required to deliver data-as-a-service by CDOs, infrastructure and operations technology by CTOs, broader digital assets including software applications and communications platforms by CIOs couldn’t be more different than that of business domain competent business analytics experts who can deliver AI/ML, Optimization, and Measurement solutions.
The very concept of combining Data, Analytics, and AI/ML responsibilities into a single CDAIO role is itself a set up to fail. Independent C-level roles and clear division of responsibilities is much more effective, akin to many other CXO roles within an organization.
How to Improve Analytics-Value Delivery?
Business value delivery through advanced analytics is accomplished through application of decision-sciences, that improve the efficacy of decisions and minimize risk from uncertainty. Successful value delivery with advanced analytics by a CAO is largely dependent on the innovation, technical rigor, and execution competence in partnership with functional business CXOs. The most important skill of a CAO is critical thinking needed for decision-optimization. They own formulation of an analytics roadmap and oversight of scientists solving problems through statistics, artificial intelligence, optimization, and simulation algorithms that focus on decision-sciences. CAO leadership starts with deep decision-sciences knowledge, and success is built with a track record of delivery learning from many different projects run through business partnerships.
Vast majority of failures I’ve witnessed are because people leaders and general business administrators are put in-charge of advanced AI/ML and optimization analytics. Assuming data-as-a-service is cost-effectively delivered by IT/OT, it is crucial for subsequent value-chain delivery on research, discovery, formulations, designs, model development, validation, experimentation / testing, measurement, and close partnerships and collaboration with business function leaders is also done effectively. CAOs should own decision-sciences and provide guidance on risk and reward through analytics and provide specific recommendations to Business leaders. Together, all CXOs should champion IT/OT as a data enablement function and allow CAOs to own Decisions-Optimization.
References
- Why Chief Data and AI Officers are Set Up to Fail, 06/2023, Randy Bean and Allison Sagraves
- “Data silos are the greatest stumbling block to an effective use of firms’ data”, 08/2018, Gaurav Dhillon
- “2024 CDO Insights: Data & Generative AI”
- “Data Quality Crisis: New Survey Reveals”, 06/2022, Elliot Leavy
- “Suddenly Doom & Gloom?” 2023, Korn Ferry
- “Six Organizing Models for Analytics Teams”, IIA
