If you think you need a Chief Analytics Officer, what you really need is a Chief Decision Officer
I saw this article in DM Review today - Building the Analytic Organization - in which Peter Graham lays out a case for a Chief Analytics Officer and how that role would be part of improving the use of analytics in an organization. Peter's right that the use of analytics is evolving and maturing in organizations but what I think he misses is that the purpose of analytics is to make better decisions, not do better analysis. While analytics are crucial to making better decisions, I think the focus on analysis and analytics implicit in the title Chief Analytics Officer is unhelpful. Strategic decisions require one kind of analytics (reporting and ad-hoc query), operation decisions another (executable predictive analytics) and some none at all (regulatory or expert decisions based only on rules). A Chief Decision Officer would be focused on helping the organization make better decisions. This would include process, people, data, analytics, rules - everything that contributes to better decision making.
Peter's stages of analytic development do resonate with me somewhat, although his focus is purely it seems to me on knowledge workers only - on delivering "analysis" to different parts of the business. His list of analytic delivery vehicles, for instance, is completely focused on knowledge workers - "static reports, dashboards, scorecards, guided analysis, alerting, and ad hoc analysis". Given this list I have to assume that when he mentions predictive analytics in the article, he is really talking about what I call predictive reporting - no executable analytic models here.... This lack of focus on executable analytics also means that he does not address the fundamentally different needs of reporting/analysis and prediction/action in terms of the data needed. For instance data warehouses and reporting often summarize data by time while predictive models often need the original time sensitive data turned into "days before event X" for analysis. Any discussion of data infrastructure should take this into account.
I believe that the next level of analytic value comes from focusing on operational decisions and applying the right approach to managing and improving them over time (enterprise decision management or EDM) and on using adaptive control for constant, systematic improvement. Peter completely misses the need for a decisioning platform that works with automated systems - how can you deploy analytics "on an industrial scale" as Tom Davenport says without one? You need a way to push your analytics into decision services, a decision service hub. After all those who act (correctly) first, not those who know first, win. Acting first means having your systems respond first and that means using EDM.
Like one of my readers, I believe that a centralized analytics group is a consequence of widespread use not a driver of it (see this discussion) and that a premature drive to a central analytics function is a consultant viewpoint. While I think companies should consider a VP of decision management or a Chief Decision Officer, I think this role is much more of a coordinator and enabler than a manager of large numbers of resources. If you are wondering if you are ready for this kind of thing, check out there posts on readiness that I wrote on my other blog:
Lastly Peter talks about competing on analytics without mentioning or crediting Tom Davenport's book Competing on analytics (reviewed here). Tom also wrote a nice piece on decision automation that he references on his blog in his post on automated decisions sneaking up and that I blogged about here.
Technorati Tags: adaptive control, analytic application, analytics, automated decision making, BI, business rules, competing on analytics, data mining, decision service, decision service hub, decision-centric, EDM, enterprise decision management, evidence-based, predictive analytics, business intelligence, Tom Davenport