Serious Thinking
Managing change in the era of Prescriptive Analytics
By Bernardo Almada Lobo*
Analytics can be defined in different ways. One of the most popular definitions considers that analytics is a scientific process of transforming data into knowledge to promote better decision-making. It is an end-to-end process that involves several steps, starting with the framework of the business problem, and ending with the management of the life cycle of the model that has been specified, developed, validated and implemented. An analytics project does not have to necessarily involve a system embedded in an automated process (with greater or smaller interference from decision agents); it can simply be a one-off analysis of the design or an improvement of a process or service.
Even though more technical tasks are important – such as identifying the analytics problem, collecting and handling the data, selecting the methodology (approach) and creating the model –, defining the initial business case and the change management when adopting the solutions prescribed by analytical methods is key to obtain the desired return on this type of project. Change management should be understood as a process of transition from a current state (for example, without an analytics model) to a future desired state (with a new model).
Most organisations (80%) already use (more or less effectively) descriptive analytics to understand past events, while a smaller number (27%) use predictive analytics to anticipate scenarios and to estimate trends, and only a minority (15%) trigger optimum or intelligent recommendations, and simulate the future results of decisions through prescriptive analytics (which requires the use of the other two types of analytics in advance). One of the reasons for this low percentage is the number of unsuccessful prescriptive analytics projects, which should not be disregarded. The necessary change in the mind-set of companies to use optimisation models and business decision support systems requires more than just technology, people and appropriate processes.
For a successful prescriptive analytics projects, there should be a buy-in of all stakeholders involved, from top management (their commitment and enthusiasm must be visible), to the holders of the data, collaborators affected by the change, and the end users of the solutions and new analytics processes. An individual’s response to change is the result of their personality and interaction with environmental factors (such as pressure to change, effective communication regarding the change and potential impact). The involvement of stakeholders in the change process mitigates the resistance to change, thus increasing the probability of it being accepted and implemented. The way the commitment between people and the change process are built and maintained goes through several stages to support the adoption and internalisation of analytics solutions.
Of the two types of analytics projects mentioned previously, the implementation of decision aid systems (SAD) is more demanding when it comes to change management processes. For some stakeholders, experts in a certain area, the new analytics process can trigger the perception of risk in their own position. In the SAD, it is key to put an emphasis on the potential for the system to explore different scenarios in a short time interval, and not on the optimality of the results generated. The SAD cannot be a black-box, but a glass-box. The users should guide the decision process, interacting with the system, sometimes conditioning the analytical recommendations of the SAD (when, for example, the user holds privileged information that has not been shared previously). The process should be transparent, simple, robust, complete, and incorporate critical business requirements.
Additionally, it is important to size the teams for the transition when companies initiate change processes based on analytical methods. In these cases, not only is it necessary to assure the business-as-usual, but there is also an effort from managers and collaborators to implement the changes to the ongoing processes or systems. The change process should be short; if that is not possible, the milestones should be closely monitored. It is very important to define a set of metrics that demonstrate the success of the process’s implementation, and their impact comparatively to the initial situation.
Other than the aspects concerning change management, the specific nature of the techniques and work to be developed, in particular the uncertainty regarding the approach to use (there is always a dose of R&D tasks), the availability and quality of the data, and the knowledge obtained with the models make analytics projects a unique challenge to all optimisation enthusiasts.
*Member of the Board of INESC TEC, Professor at FEUP
Bibliography
Davenport, T., Kirby, J., Only Humans Need Apply: Winners and Losers in the Age of Smart Machines Hardcover, HarperBusiness, 1st ed., 2016
Levasseur, R.E., Building Analytics Decision Models That Managers Use - A Change Management Perspective, Interfaces, vol:45(4), pp.363-364, 2015
Lustig, I., Balaporia, Z., Kempf, K., Milne, J. Saxena, R., Change Management for Analytics Projects, Informs Conference on Business Analytics & Operations Research, Orlando, 2016
Sirkin, H.L., Keenan P., Jackson, A., The Hard Side of Change Management, Harvard Business Review, October issue, 2005
