Operationalizing Analytics – How to Move from Projects to Production

No revolution ever happened with the support of the establishment.

—Jennifer H. Woulfe

Advanced analytics and AI projects and efforts can become exceptionally long term and conceptual if not designed, directed, and managed appropriately. I have always had a very pragmatic and practical view of working with data and analytics. Managers and end users are typically impatient when seeking numbers and indications of performance or improvements; it helps to keep their impatience in mind when executing analytics projects.

In this chapter, we will examine the process of moving from a program orientation or a project focus to a production focus. This is where we are moving the results or findings from models, applications, and analytical work products into production systems. This move sounds quite clear cut and simple when described in this manner, but there are several considerations...

The change management process

I am of the opinion that we, as analytics professionals, are seeking to drive improvements and changes in operational behavior, processes, and metrics. We are not undertaking data and analytics projects to indulge our intellectual curiosity; we are doing so to drive change and improvement.

When we are talking about driving change in a business operation, academic institution, government agency, or other environment we need to consider the entire process and problem; from the first step of the definition of the challenge to be addressed to the last step taken in proving that the modified functional system(s) and process(es) are operating in an improved and improving manner.

Figure 8.1: A comprehensive view of the change management process

So far in this book, we have discussed a number of salient factors to consider when building an analytics function in an organization, including the relationship between the analytics team and executives...

Getting to know the business

If the business is a complex operation like biopharmaceuticals, the time required for the data scientists to learn the business cannot be underestimated. Also, the time to discover the strategic areas of interest and importance of the business must be uncovered and the intersection between strategic importance and sponsor interest must be discovered as well. This is a multi-layered information and organizational ecosystem that only becomes clear with engagement and personal interaction. Understanding this ecosystem and discovery process takes time to develop, decompose, and internalize.

Change management

An aspect of analytics that is routinely ignored or overlooked is the leveraging of results or insights to deliver an improvement in operations either as a one-time modification to a process or as a permanent change on a continual basis. Implied in any process of improvement is change. One of the hardest things for people to understand, plan for, internalize, accept, and benefit from, is change.

Change of all types presents challenges for managers and for staff members. Leaving the realization that change is a required element of a complete and successful analytics project as an implied aspect of the project rather than making it an explicit agreement within the project framework leaves open the possibility for a significant mismatch in expectations between the analytics team and the functional team. All teams that need to be involved in implementing the many incremental changes required to realize the value of the overall analytics project need to be explicitly...

Analytics and discovery

When you are building an analytics team or capability in an organization that has little to no experience with data and analytics, the expectation from many people, including executive management, will be that the analytics team will discover new insights about the business; this is not an unreasonable expectation.

Given the fact that you and your analytics team will be given the chance to look across much, if not all, of the organization, you will have access to nearly all the data available internally and externally, and you will have the goodwill and collaboration of a substantial number of sponsors and subject matter experts, as well as external partners and vendors, you and your team should be able to discover new insights.

This type of discovery-oriented analytical work is typically driven by the art of the possible. You can plan projects, but discoveries rarely show up when planned. Your team may find interesting insights a few days into the...

Analytical and production cycles and systems – initial projects

In the cases where the analytics team is simply looking to improve a well-known process (that is, a non-discovery oriented improvement process), the analytics team should be able to give the functional team, sponsors, and interested executives a fairly reliable estimate of when the analytics process will start and produce the desired model, application, or result. In the first iteration of approaching a business challenge, we should have a reasonably accurate sense of how long the complete analytical project and cycle will take.

Figure 8.3: The analytics cycle

In contrast, as we discussed above, discovery-oriented processes are different and will take more time, and it is notoriously difficult to estimate and predict when the discovery will be unearthed.

Also, as we discussed in the finance scenario outlined above, we need to be clear with our collaborators from functional teams that their deadlines...

Summary

As noted at the beginning of this chapter, moving from a project or program perspective to production operations is a critical juncture and transition in every analytics effort. This transition is the point in the analytics process where the majority of projects fail. This level of failure is surprising and disappointing to me personally. I have been vexed by this hurdle and have spent years looking for ways to increase the rate at which analytics projects become successful business projects through skillful navigation of this transition with our functional managers and subject matter expert collaborators.

Analytics professionals need to grasp the importance of successfully completing this transition. This transition is critical to achieving value realization. We, as analytics professionals, need to be aware that our projects, beyond finding interesting insights and information, only deliver value if the models are moved into production and help drive improvements in operations...