Realizing the considerable benefits that prescriptive analytics can bring to an organization requires that several issues are adequately addressed. These include technical, organizational and operational factors and include:
- Large, complex optimization problems require a team of professionals trained in operations research or some related discipline. Some smaller problems can be addressed by non-specialists using technology such as Excel Solver, but most real-life problems will need an experienced team and significantly more sophisticated technology.
- High level sponsorship is needed, since prescriptive analytics usually span functional silos. The optimal solution to enterprise problems may seem sub-optimal at the department level for example, and so there will have to be mechanisms put in place to allow such issues to be resolved.
- The technology platform must scale and offer high levels of performance. While initial projects may be comparatively modest the scale and scope will rapidly grow as benefits are realized. Performance and scaling bottlenecks will be experienced if the supporting technology is architecturally weak.
- Integration with existing analytics tools and business applications means inefficiencies can be kept to a minimum and errors largely eradicated. Business intelligence, predictive analytics, rules based systems and some transactional applications will need to be integrated with the prescriptive analytics platform. Unless this can be achieved the speed and accuracy of optimization will be compromised.
- Obviously there needs to be adequate monitoring and management of prescriptive analytics projects, with effective reporting mechanisms so that changes in the business environment can be responded to in a adequate manner, and changes in business strategy quickly implemented.
It should be clear that prescriptive analytics is deeply concerned with the operational efficiency of an organization and needs to be integrated into the information systems environment. Supporting information needs to be extracted from other systems and sent to operational systems to implement the resulting solutions. Parallel with this is the need for management and reporting structures so that associated issues can be resolved. Without this enterprise support the prescriptive analytics efforts tend to remain isolated and inefficient.
Finally it is necessary that the whole prescriptive analytics effort is business driven, with a good understanding of where the major payoffs are and how projects should be prioritized. For organizations inexperienced in the domain this may mean using external resources (such as consultants and experienced suppliers) to formulate a strategy. In some industries it may be possible to buy solutions to specific problems, and inevitably the options here will grow rapidly over coming years. However it really is very important that organizations do not end up with multiple point solutions, and worse still with solutions that will not scale. And so the issues listed above are just as applicable to solutions as they are to deploying a prescriptive analytics platform.
