Whatever happened to the CIO in the age of data analytics?

I was attending the Monday morning staff meeting with my boss, the GM, along with my functional peers.

"Why is our risk exposure so high in Asia?" asked our GM.

Each of the functional heads gave an answer, mostly based on experience, gut instinct, and vague recollection of facts.

"I really need to get an analysis of our installed base by financial product, tenure, size of client, country, and any other relevant factors you guys can think of. We need to understand the trend of the risk exposure in Asia over the past three months. What caused us to close more deals and push us over the threshold of risk?"

In other words, each of our financial deals had passed the credit risk filter, but in the aggregate, exceeded our geographical risk threshold.

The GM turned to me, the CIO.

"Can you collect the data and have your guys do a time-series graph with filters so we can look at it in different ways?" he asked me. "Also, do a Pareto analysis to figure out our risk exposure by customer segments."

"Sure," I said. "Give me a week or two. You know how it is. Different CRM systems, different semantic conventions, bad data, and all that."

This scenario played out pretty much every week with different metrics and issues.

A CIO - a Chief Information Officer - was expected to know or get this information quickly and in time for making decisions. 

So why do we need a Chief Data Officer?

CIOs have always been saddled with legacy applications and development projects. CIOs should really be called Chief Applications Officer or Chief Programming Officer (though that sounds like a media network job). Someone has to lead these efforts, which takes the traditional CIO away from the specialized skills required to deal with the explosion of data.

This isn't necessarily bad. If anything, this movement to hire or place people in CDO roles brings additional and much-needed focus on the importance of data, which is to say, information.

This increased focus also highlights the need for very specialized skill sets in data science, Big Data, statistics, and machine learning. The CDO needs to have some experience or knowledge of these subjects, but the CDO cannot be the hands-on scientist.

Those who think a statistician or expert R programmer can become a CDO are making the same mistake as those who think a master C++ programmer can become an enterprise architect.

The call for CDOs is the realization that there's more to information than the traditional applications and databases.

Could a CIO become a CDO or even absorb the CDO's responsibilities?  Sure. But the CIO's organization needs to be designed right to ensure that many other areas get the focus they deserve.

Indeed, after the initial wave of hiring technical data scientists (a necessary step, of course), companies are now realizing that the most important 'language' that data scientists need to know is the language of business. 

A solid focus on business problems, priorities, and insights is necessary to make the analytic function sustainable and continually relevant.

A data scientist must know how to interpret the results of analysis, package them for consumption by non-data scientists, convey the limitations of models, discuss possible biases (in the vein of Kahneman, Slovic, and Tversky), pros and cons of techniques, associated risks, and so on.

Then, there are all the usual messy data problems that lay professionals don't readily associate with data science: metadata, data quality, data governance, data warehouse, reporting, business intelligence, etc.

To get true business insights, you have to look at information that is generated from the interactions of the company with its external constituents such as customers, partners, suppliers, and regulators.

To make those insights actionable and effective, you have to work with data inside the company.

That, at the senior-most level, is truly the job of the CDO.

I see the Chief Data Scientist as the senior-most analytics technician (probably the coolest job right now!), while the CDO is the executive role, just as the CTO is the senior-most IT technician while the CIO is the executive/ managerial/ administrative head.

I think it is important to clearly define these roles and understand the competencies required in each. Otherwise, we are in danger of asking the CDO to write R programs or asking a data scientist to own the budget for analytics.

We have enough problems to deal with. We don't need the additional confusion of poorly defined expectations of the CIO/CTO/CDO/CDS roles.