“Without data, you’re just another person with an opinion.”
- W. Edwards Deming
Who knew that I could find a quote from W. Edwards Deming on data? Never mind that he was able to produce a massive body of work that altered numerous industries the man was a quote generating machine. And he was prescient in his use of statistical methods to drive business change. I read his works over thirty years ago and I am still amazed at how much of what he wrote is relevant in a world that is vastly different than his time.
So back to the title of this article. How do we define what a CDO does? Good luck with that one. Every company has their own definition. Often it feels like a dumping ground of the sort of feel good, should get done but will never get done, looking for problems to solve kind of role. As part of doing some background research before I wrote this article, I thought it was wise to read what some of the leading management thought leaders wrote about the role. The role has been around for several decades, but it is still poorly defined. That’s management parlance for its evolving, and nobody knows what you do. As an outsider looking at some firms that have hired CDOs I am often bemused by the types of individuals that they put into the role. You just know that the person they put there are doomed to fail. And if you read the remit they are given you just have to feel bad for the person trying in good faith to do the job. A job that is nearly impossible to be successful at because it is poorly defined, they don’t have the right tools, their role overlaps with other ‘C’ level executives, they don’t have any real engineering experience and they really don’t understand data. They might have great marketing degrees or MBAs from top universities but what skills are really needed?
They get tasked with these nebulous concepts such as compliance, data governance and data quality. And the unwritten requirement is that they must institute massive cultural change. Change that can be highly disruptive without the remit to do such change. Who in their right mind would even attempt to go into such a role where failure is virtually guaranteed? Data quality is one of those concepts that I find unmeasurable. There are those that will swear by it but, and call me crazy, but if your systems are operating with really bad data quality, you have much bigger problems than data quality issues. You have some serious foundational engineering problems. Might want to fix those first. Deming said it perfectly that you need to build quality into a product and not check for it after the fact. It’s too late by then and the cost of fixing it has gone up tremendously.
My opinion about data has always been to shed some sunshine on it by letting the business users view it and you would be surprised how fast issues get resolved. I didn’t need a separate organization and a ‘C’ level executive to do that. Data governance is another one of those nebulous concepts that everyone loves to opine about. We would all love to know where our data comes from and what is its provenance. Simple enough but in a world of super complex, highly distributed systems trying to find the path my data traversed and how many systems touched it before it got to me is no small effort and quote bluntly the merits of which are questionable. You can make the case in some domains that it’s necessary, but I have found those to be few and far between. Even data in simple business processes can have a complex provenance.
The problem with most companies’ definition of what a CDO should do is that it is so far removed from the front lines. For most companies it’s a role of influence and not delivery. The engineers are the ones that create and manage the data daily. When someone in a CDO organization comes to speak to the engineers about data quality and governance their eyes immediately glaze over. It’s not relevant for their day to day. CEOs think they need to have a CDO because that’s what the rest of the world is doing. I am Mr. Big Bank CEO, and the rest of the corporate world has a CDO, this brand new ‘C’ title so I must have one since I have petabytes of data that I can use and maybe they can figure it out. It’s not quite so simple which is why the turnover for the CDO role is absurdly high with the average tenure around two years. Now the role itself can be of immense value if it is properly defined with a clear remit.
How the role should function.
So what do you do? CDO and Chief Analytics Officers (CAO) roles should be combined into a single unified role that can have far more business impact than either one independently. CAOs struggle because they don’t have access to the data they need to add business value. CDOs get stuck doing nothing but creating fancy PowerPoints and rely on other teams to deliver anything. Together and, with the right remit and reporting lines, the consolidated role can have a tremendous impact on the organization.
Having worked with three separate data science teams I can say with certainty that those with broad access to data and can control their own data pipelines are far more successful than those that must wait on other divisions to give them access to a dataset. So much of data science is being able to test new datasets, enrich existing ones, transforming data into different formats, acquiring new datasets and the flexibility to manage their own data is critical. Firms that force their data scientists to beg for data and must wait on an engineer in a business line to give them data are wasting critical time.
The combined role of CDO and CAO needs teeth. It needs a budget along with with its own engineering staff to build out the infrastructure necessary to support the data science function. It also requires a staff of data scientists and having data engineer and data scientists work side by side creates a synergy that feeds on itself to create corporate value. A successful combined role is highly impactful to an organization, and it can rapidly monetize the firm’s data assets. In many cases the CAO is already doing most of what the CDO is tasked so why not further augment the role, give it more teeth and budget and have it be delivery focused?
The combined role can and should report into the CEO. It is not a technology function but a role that helps every aspect of the business from strategy, product and customer optimization, HR functions along with every other function in the business. Reporting to the CEO gives it the necessary gravitas so that people listen. It can and should be a profit generating division. All firms need to monetize their data. I don’t view this role as managing all the core data infrastructure in the firm and that should be left to the CIO, but there are many tools to facilitate access and combining various datasets easily. But it is critical that this organization have its own set of infrastructure and tooling for data management, visualization, and analysis. Functions such as compliance, regulatory and quality should be pushed to the heads of compliance and are not part of this role.
The person in the combined role should have deep engineering and analytical skills, soft skills, excellent communication, leadership, and strategy skills. They need to operate at the ‘C’ level comfortably and become a core part of the management team.
Thus, the combined role with the proper support and budget can be hugely impactful for an organization. Far more so than keeping them separate roles.
Much like the old engineering KISS principle keeping the organizational structure simple pays long term dividends.