In part one of this article, we discussed how predictive analytics can be used to help predict the future, providing a crystal ball that can help organizations predict how to better reduce churn by addressing customer needs, add relevancy to selling, or more accurately assess risk. Predictive analytics provide the pieces of the puzzle that produce the recommendations that come out of the crystal ball. Now that we have an understanding of what predictive analytics is, let’s look at how it can help today’s businesses with a crystal ball that helps drive growth.
Predicting the analytics evolution
Predictive analytics is no longer exclusive to large enterprises, but many of these global operations have been able to take their use of predictive analytics to the next level simply because they are more mature in their use of the technology and started managing their data earlier. But predictive models are pretty useless by themselves. Today’s organizations need to be able to act on the insight provided by predictive analytics. Evolving predictive analytics into a business’s underlying processes gives the organization the power to predict and then act on a recommended next-best-action. The ability to do this repetitively allows for an always-on customer decision hub, or recommendation engine, that uses insight into expected customer behavior to increase lifetime value.
The big picture
Taking these real-time recommendations is the next step in the use of many predictive models. Participants leverage these predictive models in the context of their work, and make the processes pro-active and dynamic, for instance with prioritized recommendations for optimized guidance to employees when responding to customers or resolving their tasks. This ability is not limited to employees in contact centers, as the same recommendations can be used in self-service channels to anticipate customer needs aspirations and balance them with business objectives. Leveraging this insight during the moment of truth is the driving force behind customer differentiation. In a typical business 80% of the profits come from 20% of the customers. As a consequence, every interaction needs to be informed about what resources can be justified to improve the customer experience. By leveraging a ‘sneak preview’ into the future, organizations are able to not only serve their customers better, but they are better prepared to help drive corporate growth through much better targeted services and offers.
The need to combine this ability to predict with underlying processes can be best demonstrated by an offer and order fulfillment process. If you contact your mobile phone provider and during the course of the interaction, they are able to present you with a relevant offer (through their use of predictive analytics) that is of value to you and you decide to accept, the next step in this process is finalizing your order and fulfilling your request. It could be that you decided to upgrade to a new mobile device. Once you accept the offer based on the predictive model, it now needs to be handed off and fulfilled. If an organization has not unified their predictive capabilities with their operational processes, this hand-off becomes much more challenging. If these two capabilities are unified, then we can have an effective process that delights the customer and also provides the mobile provider with growth.
While unified, there is still a necessary separation of concerns. Those that design an end to end loan origination process for instance are almost certainly not the same people that design the decision strategy that decides on who should get a loan in the first place and whose application should be rejected. So while, for a seamless experience, predictive models and prediction-driven decisions should ideally execute on the same platform as the processes they’re embedded into, the design phase should be separate, the process describing how to do things, the decision strategy describing what to do.
Mind the data
One challenge and word of caution that all businesses need to adhere to when using predictive analytics is that predictive models need to continually be monitored so that models do not become ‘tired’ particularly since the volume of global data is predicted to expand by a factor of 44 from 2009 to 2020 and reach 35.2 zettabytes, according to IDC. If establishing this monitoring is not a priority, the data upon which the model is built will no longer be representative of current circumstances and the recommendations that are provided become increasingly less relevant and detrimental to business growth. For example, if the customer demographics for an organization are changing, the average age of your customers could be increasing for example, and the models that are being used need to keep up with these changes. If not, you may find out that the recommendations, tailored to younger consumers, are missing the mark. And that’s just age. Models need to be continually monitored, both their input as well as their output, not just so that they can keep producing relevant results, but also so that they keep aligned with corporate objectives. If not, many organizations could see offer acceptance rates suffer as a result.
Hungry for more data
The use of predictive analytics can become quite extensive within many of today’s sophisticated, global organizations. With the ability to provide propensity, risk, and attrition models across a number of various products and services, organizations will see an ever increasing appetite for predictive analytics. Whether they realize it or not, many organizations require hundreds or even thousands of predictive models over the course of a year, particularly with the amount of change industries are faced with every year. This is where the need to shift from a statistical laboratory to a fit-for-purpose predictive model factory becomes evident. Such a factory provides the capability to safely produce a steady stream of models at the pace of business. Over time, organizations may also see the need for models to continuously adapt and self-learn. This would involve bringing adaptive analytics into the fold (which will be covered in a future edition).
Predictive analytics have become the life-blood of the real-time enterprise. Because of the ability to more easily develop reliable models, predict valid outcomes and drive business growth, today’s enterprises view it as a critical capability to help them survive. Because of this, predictive analytics have matured from a “dark art” to a mature function that helps to fuel today’s businesses in piecing together the puzzle to better understand and serve their customers.