Routes to the Top - Data Science & Analytics

​The roles of Chief Data Officer (CDO), Chief Analytics Officer (CAO) and Chief Data & Analytics Officer (CDAO) are relatively new, having grown in adoption significantly over the past decade as data has become ‘the new oil’ and as traditional companies seek to contend with digital competitors. These senior roles are now found within a variety of data-intensive industries, especially within Financial Services and Healthcare, and increasingly in others too.

...There is a wide-open field in many industries for developing products and services that have data and analytics baked into them.” 
TOM DAVENPORT, Information Technology Professor

​A 2019 NewVantage Partners survey found that almost 70% of respondents reported having a CDO, significant growth from 12% in 2012 when their first survey was done. While Financial Services firms compromised the majority of respondents (“data-mature firms that maintain high-value customer relationships and invest significant sums to manage their data”), Healthcare firms also shared their views (“businesses that are undergoing rapid transformation – data rich, but often less data-mature”), as did businesses in other industries. The survey highlighted the “creation of a yet newer role [than CDO] – the Chief Data and Analytics Officer – integrating the data and analytics responsibilities within a single executive.”

As Forbes Insights puts it: ““It’s safe to say that a flood of data – 250 billion terabytes created every day – has swept the chief data officer into the C-suite.”

As firms globally strive to harness the power of data and face challenges along the way, the demand for data scientists and analysts continues to increase. This post discusses the evolution of data-driven roles, some challenges of executive roles in these disciplines, and key skills to improve the likelihood of success in these roles, drawing on the insights of Tom Davenport, A. Charles Thomas and others. Read more to explore the nature of data and analytics roles, and their growing executive influence, in this latest entry in our Routes to the Top series...

C-Suite Data & Analytics roles today
The first known Chief Data Officer in the business world was appointed by Capital One in 2002:  Cathryne Clay Doss. Her responsibilities included strategic oversight of IT and supply chain and market analysis. In 2005, Yahoo became the next major company to appoint one.

Fast-forward to today, and the landscape is significantly different. Most business leaders view the CDO role as a critical role, says Forbes Insights, but “no one agrees precisely on what the role entails, including CDOs. How much technological expertise do they need, how much data governance responsibility should they have and, especially, to whom should they report are all open, and increasingly pressing, organisational issues.”

One of the factors driving companies to appoint senior executives in data and analytics roles has been the fear of disruption from nimble data-driven entrants. These emerging digital competitors may not have had the access to proprietary data though, giving established players an advantage if they could attract or develop the skills to analyse and interpret their data. Often, digital garages were set up to experiment with new ideas and technologies, separate from the entrenched processes and culture of the existing business. The NewVantage Partners Executive Survey (2019) describes it like this: “...companies have come to increasingly recognise that they must become more adept at leveraging their data assets if they are to compete successfully against highly-agile data-driven competitors.”

Given the newness of the role compared to other C-suite positions, there is a range of definitions and implementations across industries and companies. For example, some say that the primary responsibility of the role is setting the data strategy within the company and owning its results, while others say that responsibility belongs to other C-suite executives or argue that no single role is accountable for data. Less than 15% say that the role should have revenue responsibility. A majority of CDOs say they are accountable for data quality, data governance, master data management, as well as information strategy, data science and business analytics.

Tom Davenport, professor of information technology at Babson College, argues that the shifting role of the Chief Information Officer (CIO) over time accounts for the lack of consensus about the CDO role. Data governance, he says, “was previously the function of the CIO. Many organizations have felt like, despite having somebody theoretically responsible for it, they were not getting the kinds of outcomes they wanted with data so they created a new role.” Part of this is the increasing importance of fostering a culture of data-driven decision-making across the whole business, using data and insights to improve business results, operational efficiency and effectiveness.

​Richard Wendell, board member of MIT's International Society for Chief Data Officers, says that the range of responsibilities may simply reflect the different needs of businesses and how mature they are in their digital culture: “companies have different needs at different stages of their technological evolution” i.e. not one-size-fits-all. To him, a fundamental aspect of the role is leveraging the company’s data as a strategic asset.

Today’s senior data & analytics leaders reflect diverse backgrounds, representing a mixture of company or industry veterans and external change agents, and different histories: data scientists, business executives, or technology experts. This diversity is also likely to be a consequence of the scarcity of these leaders, allowing them to move across industry verticals more easily. It also means that those aspiring to these roles can take different routes to get there.

Evolution of a Data Scientist
While a strong CDAO will have a blend of skills across data science and analytics, technology, and business savvy, they are more likely to have started from a quantitative, analytical, or technological background, and added business understanding through work experience or further study (such as an MBA).

Johanna Pung made this for Wikimedia Deutschland [CC BY-SA 3.0]

An illustrative route someone with strong analytical skills may have taken to reach a CDAO role is:

• Gaining quantitative skills through studying for an analytical, statistics, finance or technology degree
• Work experience building analytical models and solving defined problems for a business area that recognised the commercial value in data analysis, such as a marketing department wanting to determine which customer needs to target or assess the impact of specific marketing initiatives
• Their initial role(s) would have been as an individual contributor / manager of self working with transactional data (e.g. marketing, credit and retail data), and they’d be seen as “the person with the answers”
• Over time, as they built wider relationships, the value of their analysis would have become more widely known, and they’d work on bigger challenges, perhaps reporting into a Customer Experience function
• Those with leadership potential and interest would step up into managing others, steering a team of analysts to solve problems within their line of business. Under their guidance as a Manager of Others or Manager of Managers, their team(s) would develop technological solutions that enabled colleagues to run their own reports and search available data for insights i.e. dashboards or self-service business intelligence, partly as a response to increasing demand for the valuable analysis outputs​ 

• As the demand for their teams’ data-driven insights grew, an opportunity to take on enterprise accountability for data and analytics may have arisen, including working towards alignment of data, methods, techniques, tools and governance across business-unit teams to improve consistency (for example by providing a singular view of each customer), democratise data access (through a “data warehouse” or “data lake”) and manage risk. In this role, they would be impacting every function of the company, ultimately operating as a Manager of Function themselves: the data and analytics function, recognising that this can have varying degrees of centralisation
• In some companies, the data/analytics function can become a business in its own right, when it succeeds in generating its own revenue (i.e. P&L responsibility). In this case, the head of the function operates as a Manager of Business, acting as CEO of a data company within a larger corporate entity.

​This evolution very roughly mirrors the career of A. Charles Thomas, General Motors’ first-ever Chief Data and Analytics Officer (CDAO), which he describes in his keynote at the Wharton Customer Analytics conference. Charles has also worked for Hewlett Packard (HP), the United Services Automobile Association (USAA) and Wells Fargo, and he discusses his own shift “from back office to front office to running a company” over his career. He explores how analysts more generally began by cranking out the data, before slowly getting into managerial positions, and more recently into the C-suite. 

It’s roughly 1 hour long, including a Q&A session for about half of this.

Over my own career, I’ve seen a similar evolution play out for other major trends too, such as the rise of asset-liability management (including hedging and risk management) within an insurance context as financial economics skills entered established businesses. 

Charles describes his early experiences of how data was originally overseen by the IT function, with analysis being done in another area, with little communication between the two. The CDO and CAO functions tended to be separate (especially in Financial Services), with the CDO handling governance and the CAO analytics. As CDAO he now has singular accountability, although he does work closely with the Chief Information Security Officer and the Chief Privacy Officer, especially on matters relating to privacy and hacking. Now, in what he sees as CDAO 2.0, he views himself as CEO of a data company, monetising data-as-a-service.

Interestingly, this iterative evolution of the role of data scientists and analysts is reminiscent of the iterative nature of data science itself. Kanika Bhalla describes it as “an iterative process, with opportunistic improvements being made to data, features, models and visualisations throughout a project’s duration.”

Challenges of the executive-level role
The lack of consensus at the executive table about the precise remit of a CDO, CAO or CDAO can make the role particularly challenging. Different expectations can lead to severe disappointment for many, especially if some believe that data will be a silver bullet that will solve any company problem.

It is crucial to be clear on expectations as soon as possible, and communicate what your short-term objectives are as often as you can. For example, you may wish to unlock powerful data for use across the enterprise. But, be careful: in Tom’s words “data is a mess in almost all organisations”, so being seen as the person who will fix it all is a recipe for disillusionment. Along these lines,   Charles explains that your job is to retrofit the technology of the company, given there is a myriad of disconnected technologies existing already.

The CDO role was a good idea, but it very much needs role clarity and the ability to provide rapid value. The fact that data is a mess in almost all organisations would seem to provide some job security for CDOs, but it doesn’t.” 
TOM DAVENPORT

Tom also explains that “effectively managing all data across a large organisation is too difficult for anyone to pull off, and the job needs to be more narrowly specified if its incumbents are to succeed.” Defining the problem(s) you’re trying to solve is good advice. 

As you’re acting as a change agent, it is likely you would need to achieve things by working with existing structures in the company, using governance and influence mechanisms to implement improvements, rather than formal reporting relationships. Richard explains: “Being a change agent and a strong communicator is really important. [You must] be able to win the political battles and make data available throughout the organisation. The main reason why companies want a CDO is because people are frustrated about not being able to access the data they need to do their jobs.”

​Other techniques to help change the culture include:

• multiplying your business impact by shifting power or access into the hands of users by democratising data so that everyone can run their own analysis to address their pressing questions through self-service business intelligence – Charles refers to these empowered advocates as “citizen data scientists”
• winning the confidence of others who may not have deep technological expertise, including others at the executive table
• working hard to change attitudes and behaviours around data, recognising that internal resistance is normal and change is difficult

Another challenge is the extent of senior support the role may have. For example, it may report into other senior executives rather than the CEO, or may be part of the second tier of the C-suite. Effectiveness depends fundamentally on the CDAO’s ability to wield influence and build support among decision makers. One potential conflict is with CIOs who often desire to minimise IT costs, rather than invest in new tools to maximise the value of data.

​Reaching an appropriate balance for the organisation between the compliance and commercial aspects of the role can also be immensely challenging, especially where other functions share some of these responsibilities. A compliance focus is a defensive one, preventing breaches, security and privacy issues and data quality problems, and streamlining data environments. In other words, a governance focus aimed at avoiding the downside of data. In contrast, a commercial focus tries to achieve upside through better customer understanding and relationships e.g. marketing initiatives, underscoring why analysis frequently begins in marketing, customer experience or other externally-facing functions. Both aspects are important, but require different skill sets and actions.

While cybersecurity and data privacy are essential (and no-one wants breaches or errors on their watch), Tom warns that if you only work on improving the internal data environment or preventing problems, “nobody on the business side will find your job, and the associated data models, data governance structures, and data catastrophes supposedly averted, very compelling. They’ll mostly become frustrated, as will you, with the slow pace of data improvement.”

If you can improve business decisions or operational performance or customer satisfaction through analytics, machine learning, intelligent agents, and the like, everyone will be much more impressed with your work. Even if all the analysts and data scientists don’t all report to you, at least try to own an analytics/AI centre of excellence or something similar. You need to get some credit for the magic that analytics and AI can bring.”
TOM DAVENPORT

I’ve experienced this compliance vs commercial balance during my career too, along with the opportunities and challenges of a hub approach in a global organisation when I relocated to establish a group function that shared best practice across countries and built new revenue streams (commercial) as well as ensured adherence to group standards (compliance). In my experience, our responsibility for sign-off gave the local businesses a compelling reason to engage with the hub early on, with the resulting collaboration improving the ultimate outcomes. 

A related area is monetisation: if you can use your company’s data to add new revenue streams by creating “data products”, the value of data and analytics becomes much more apparent.

Key skills
There is a wide range of skills necessary to be successful over a data or analytics career, stretching from analyst or data scientist to Chief Data & Analytics Officer and beyond. By their nature, they are business roles grounded in quantitative and technological expertise, so some of the skills are technical, and others are broader, helping prepare you to deal with the challenges of an executive-level role. 

The technical ones are likely to be impacted by technology and automation, so exact languages or systems can shift reasonably quickly. Some examples of technical skills are: 

• mathematics and statistics, including key metrics and distributions, Bayesian analysis, hypothesis testing, differential calculus and linear algebra
• appropriate data-visualisation (including enabling others to do this themselves) in order to explore data, as well as summarise and communicate your analysis more easily
• machine learning, both non-parametric and semi-parametric
• computer science and information technology, including the philosophy behind programming, data structures and algorithms, relational database use and management, parallelisation and distribution of large models, and performance analysis of code and processes
• domain expertise, to understand the underlying dynamics, forces and trends of the specific industry

These skills are not everything though: Charles calls R, Python and other techniques “the price of admission”. Communication skill is vital, including conveying clearly the nature and value of data and analytics. As is explaining your methods, findings and recommendations intuitively to a range of audiences from non-technical managers, to regulators, to end-users of the analysis.

The person who understands data and can explain it to the business, that’s the unicorn. That’s the hardest skill to hire for… Getting someone who gets the business, gets the operations, and understands insights is a very rare skill.”
A. CHARLES TAYLOR, Chief Data & Analytics Officer at General Motors

Charles emphasises that you can differentiate yourself by mastering communications and storytelling. This includes transforming data for operational purposes into data for insight purposes, and making your messages relevant to the audience in creative ways so that they take action i.e. influencing them. This helps you to become a trusted advisor, supporting smarter and faster decision-making. Important elements are:

• framing the business problem, 
• determining the data needed, 
• assessing what analysis will answer the question(s), and 
• acting on the results of the analysis.

He says, rather than simply responding to requests, “be out in front, leading the conversation about what’s next”. Value arises when you understand the organisational processes (e.g. HR, logistics, facilities management) and plug into them to inject analytics at the right place and time.

The skills of convincing and influencing others and leading change become ever more important as you progress in your career. Influencing skill is particularly important in cross-functional roles, and helping others be successful can encourage their ongoing support. Charles makes the case that a blend of analytical skills, bedside manner / EQ, and business-mindedness is needed.

[Be] strategic about how to obtain a reputation throughout your career. Decide how you want to be known, and work towards that.”
A. CHARLES TAYLOR

​Another piece of advice Charles gives for those in mid-to-senior roles is to ask your seniors to ask questions in the meetings and forums they attend. When they ask you “what can I do to help you?”, Charles advises: don’t ask for money or budget. Instead, he says what will help advance the cause of data and analytics across the organisation is if those leaders ask two questions whenever they see a number: (i) where did you get it? and (ii) are you sure we optimised it? This effectively directs people across the organisation to ask for your input on analysis.

Given that data and analytics can seem a nebulous mystery to many, another important skill is delivery i.e. deliver tangible, measurable and visible value as often as possible. Tom advises that “every project should last no more than a month and produce measurable results, ideally in hard currency”. He also suggests picking a data domain that is particularly important to your company, and working to make it noticeably better - “a domain that is closely aligned to your strategy and organisational mission”. 

Richard shares a similar tip for those in a new CDO or CDAO role: “In the first 90 to 180 days, a CDO should come in to be a change agent. Win some political battles to get people to stop locking data up in their own territory and make it available to the enterprise. Often just getting data more available can be a good early success metric.”

Charles has advice for managers looking to spot and develop talent too. One of his suggestions is to rotate your team members into different business functions so that they can spend time getting to know the people, and seeing their priorities and challenges first-hand. When developing others, he recommends looking at the slope, rather than the intercept: their rate of growth relies on how interested and effective they are in learning new things.

Good people will overcome bad bosses, bad plans, bad data and bad technology. And so I realised that growing people in this space was the way to drive the most amount of change. And trying to influence people to make a bigger difference as opposed to who can write the best code. ...Inspiring people to help them see the possibilities of what their career can do.”
A. CHARLES TAYLOR

Data Ethics, Privacy & Duty of Care
While the world of data and analytics has a lot to offer organisations and customers (including increasing personalisation), there remain a number of questions to address, partly driven by the different incentives and awareness of those involved. They include our nagging questions as we hear some of the ambitions of companies (such as tracking which radio station you’re listening to, or which shopping centres you visit in which order, so that they can sell this information...)

To guide this long-form post to a conclusion, I would argue that Data & Analytics leaders of the future will need to play an active role in exploring these questions and understanding their implications for all parties. For example, in an insurance context, this includes how far personalisation should extend: at the extreme, price optimisation would lead to no cross-subsidy, effectively excluding certain individuals from insurance coverage.

Some of the areas where acceptable practice is not yet coalesced include:

• Data ethics – what uses of data are ethical?
• Privacy, including opt-ins, opt-outs, and sale of personal data
• Education and awareness of what data is being gathered by devices, including cars, wearables, phones, pacemakers etc...

Regulation such as GDPR is helping to protect consumer interests and improve transparency around the use of personal data, and some professional bodies have begun exploring the questions around ethics and privacy. For example, the Royal Statistical Society and the Institute and Faculty of Actuaries recently collaborated on A Guide for Ethical Data Science.

While regulation and standards can help to protect consumers, it is not sufficient for practitioners to wait until legislators set the minimum requirements that must be adhered to. Instead, to earn the trust of customers and the public, I believe that data and analytics leaders will need to proactively demonstrate a duty of care, such as a professional responsibility to act in the public interest. Leaders who can act in the interests of both their customers and their organisations will be the ones to look up to, and I encourage you champion this on your own route to the C-suite.