Three steps can reduce churn in your company's data leadership role(s).
I recently came across this article in Harvard Business Review on the high churn in the CDO role.
(For brevity, I'll use "CDO" as an umbrella term for "the leadership-level role that focuses on company-wide matters around data, machine learning, and AI." Companies also call this the Chief Data & Analytics Officer (CDAO), Chief Analytics Officer (CAO), Head of AI, or Chief Data Scientist. Sometimes the person doesn't even have a formal title. It's simply "Jane, who brought this company out of the Dark Ages and turned all of our data into something useful." I will not, however, refer to the CDO role as the person who "owns" data. Because that's just silly.)
After describing common challenges of the role, the article offered CDOs tips to improve their chances of success. This was all solid advice. Frankly, the advice applies to data professionals at all levels of the organization, not just the executive suite. And I'm not saying that just because the advice aligns with points I have made on this website and elsewhere ...
What I found most interesting was that the tips focused on actions the newly-minted CDO could take. Given the high turnover of this role -- just two years' tenure, on average -- it's clear that the CDOs aren't the only ones who have work to do here. Confusion over the role's definition, mismatched expectations, and churn all point to a larger problem.
And that problem is, quite simply, that companies aren't doing nearly enough to position their incoming CDOs for success.
Companies can turn this around by taking just three key steps:
1/ Develop an organization-wide understanding of what data/ML/AI is and what it can really do.
This doesn't mean that everyone needs to write enough code to train and deploy a predictive model. But everyone who interacts with data, or has some decision-making power around it, needs to have enough data knowledge relevant to their role.
People in product management, for example, need to understand the typical "R&D&D" lifecycle of a predictive model. Executives should know the various ways an AI effort can fail, and how those failures can quickly translate to loss of money and PR snafus. Everyone has to understand the ethical issues around collecting, storing, and analyzing data about individuals.
2/ Based on step 1, develop concrete plans on how this company can use data/ML/AI.
Identify the specific problems data analysis or AI will solve. Then write this all down. You want to be very detailed here. "Improve decision-making through data" is too vague to be useful; but "understand usage data to better understand customer purchasing preferences" is better.
An incoming CDO will extend and modify those plans (you're hiring an executive, after all) but they'll need this as a starting point. Even with all of their knowledge of data, and their ability to quickly learn about your business model, that plan you provide will give your new CDO clear direction on your business concerns and how you're thinking about data. More importantly, it shows them that you're already thinking about this.
For bonus points, review and document all of your data sources. The incoming CDO will perform their own review, but your documentation will serve as a much-appreciated launching pad.
3/ Don't even think of hiring someone to head up the company's data efforts until you have tackled steps 1 and 2.
If you haven't mapped out what's possible with data, and what you want to achieve, then your company isn't ready to adopt data analysis and its related technologies. You're simply piling your data hopes and dreams onto the plate of someone you haven't even hired yet. Handing that incoming CDO vague ideas attached to high expectations is neither productive nor fair.
Nor are you in a position to evaluate a potential CDO hire. If you don't know what you want this person to accomplish, how can you determine they're the right fit for the job?
I'll acknowledge that these three steps are not small undertakings. They will require the time, effort, and discipline of a variety of people in your organization.
Before you try to short-change that work, just remember what hangs in the balance: the dollar, time, and effort costs of recruiting an executive, along with the opportunity costs of having to keep re-starting your company-wide data efforts. Doing your homework is the cheaper route.
(All of this is predicated on the idea that your company even needs to head down the path to data analysis, machine learning, and AI ... but that's a story for another day.)
AI-based Automation: Ways to Mix Human and Machine
Humans versus machines? To reduce your risk, the best answer is "yes."
New Radar article: "Rebranding Data"
I've published an article on O'Reilly Radar: how many times will we rename the data field?