By Dennis D. McDonald

It has become apparent to me that “data governance” can’t just focus on data and tools. 

We hear a lot these days about topics like “digital transformation,” “big data,” and “data-driven decision-making.”

A common thread is that organizations must learn how to govern, analyze, and interpret the increasing volumes of data that are relevant to them, regardless of sources or types of data. Failure to do so will, according to the pundits, result in failure, defeat, lack of competitiveness, or worse.

While I can’t argue with the importance of making sense of data, I do have a bone to pick with one of these terms, “data-driven decision-making.” What usually pops into my head when I hear this term is, “When wasn't rational decision-making not data-driven?”

Granted, I've always considered myself to be somewhat data-driven. My graduate studies were heavy on research methods, statistics, and operations research. My careers since then have focused on data intensive systems and processes, right up to my current interest in data governance.

But am I  “data driven”?

I learned long ago the limitations of data and the need for judgement in personal and family affairs. Anyone who has been faced with major life and death medical decisions will know the terror of selecting among choices with uncertain probabilities of success – or failure. Numbers don’t necessarily lie, but they don’t always tell the whole truth, either.

Consider organizations. Are organizations now really more likely to be more data-driven than they were in the past?  A lot will depend on the nature of the business and the organization. Firms that depend on high-speed augmentation of financial transaction management, for example, must seriously be considered to be data-driven, as are many firms that have grown up with the digital revolution and web based e-commerce. Calculating orbital paths and fuel consumption are also data intensive.

But what about smaller or more traditionally structured organizations that must still rely on legacy programs and processes to conduct their more uncertain day to day businesses? Is there any reason that they, too, shouldn't strive to make their decision-making more data-driven, given the increasing power and flexibility of the tools now available for making sense of data of all kinds, including data on human and consumer behavior?

Of course not. But tools of any kind need to be evaluated in terms of their costs and their benefits. Modern data management and analysis tools or no different.

How to do this? There are at least three important issues to consider when an organization strives to become more data-driven in its decision-making:

1. What to do about decision variables that are inherently difficult -- or even impossible --to quantify?
2. Even if we think we may have access to data relevant to the program or decision under consideration, are we able to clean and organize the data so it's reliable enough to support our decision-making on a sustainable basis?
3. Regardless of the amount or type of data that we can get our hands on, do we really understand how risk and uncertainty will impact the decisions we need to make?

Understanding how (1), (2), and (3) interact is not a job to be taken lightly. Nor is it a task that a single tool or data model is going to solve. Even if we have perfect and complete data that inform one particular part of our question, what if that part of the problem isn't where the most uncertainty in decision-making arises?

I think there’s a problem with the term “data driven.” While there may be a lot of problems and decisions that can be “driven” with the right descriptive, prescriptive, or predictive models, are those necessarily the most important problems or decisions that we need to address?

Perhaps it would make more sense to use the term “data informed” when it comes to making serious and complex decisions that involve significant uncertainties. We would all like to predict the future, but look at how much uncertainty and disagreement there is about what impact AI – Artificial Intelligence and its variants -- will have on employment in the future. Will AI put people out of work? Will AI create more new jobs? Will we become so dependent on AI-based services that we lose our ability to cope with real world uncertainty? No one really knows.

What it boils down to, I think, is that we need to make sure that we are focusing on the right decisions. These might not be the easiest ones to model. They might also be the ones where we have gaps in the data that will require filling with more data – or with mature and professional judgment.

It has become apparent to me that “data governance” can’t just focus on data and tools. Data governance in an organization must also take into account all the systems or processes that are relevant to a decision.  This is one of the reasons why in the past I have toyed with using the term data program management to express this complexity. As I noted in An Introduction to Data Program Management (DPM),

Data Program Management (DPM) is the intelligent application of data management tools, technologies, and processes to improve the usefulness of an organization’s data. DPM helps the organization to: 

• Improve how data are defined, organized, managed, analyzed, and used.
• Document and manage significant technical and semantic data and metadata relationships.
• Apply data assets wisely in alignment with the organization’s goals and objectives.

For me, then, the bottom line is that data doesn’t drive decisions, it’s really the other way around.

Copyright © 2017 by Dennis D. McDonald