You are currently browsing the tag archive for the ‘Predictive Analytics’ tag.
Organizations today must manage and understand a flood of information that continues to increase in volume and turn it into competitive advantage through better decision making. To do that organizations need new tools, but more importantly, the analytical process knowledge to use them well. Our benchmark research into big data and business analytics found that skills and training are substantial obstacles to using big data (for 79%) and analytics (77%) in organizations.
But proficiency around technology and even statistical knowledge are not the only capabilities needed to optimize an organization’s use of analytics. A framework that complements the traditional analytical modeling process helps ensure that analytics are used correctly and will deliver the best results. I propose the following five principles that are concerned less with technology than with people and processes. (For more detail on the final two, see my earlier perspective on business analytics.)
Ask the right questions. Without a process for getting to the right question, the one that is asked often is the wrong one, yielding results that cannot be used as intended. Getting to the right question is a matter of defining goals and terms; when this is done, the “noise” of differing meanings is reduced and people can work together efficiently. Companies talk about strategic alignment, brand loyalty, big data and analytics, to name a few, yet these terms can mean different things to different people. Take time to discuss what people really want to know; describing something in detail ensures that everyone is on the same page. Strategic listening is a critical skill, and done right it will enable the analyst to identify, craft and focus the questions that the organization needs answered through the analytic process.
Take a Bayesian perspective. Bayesian analysis, also called posterior probability analysis, starts with assuming an end probability and works backward to determine prior probabilities. In a practical sense, it’s about updating a hypothesis when given new information; it’s about taking all available information and seeing where it is convergent. Of course, the more you know about the category you’re dealing with, the easier it is to separate the wheat from the chaff in terms of valuable information. Category knowledge allows you to look at the data from a different perspective and add complex existing knowledge. This, in and of itself is a Bayesian approach, but allows the analyst to iteratively take the investigation in the right direction. Bayesian analysis has had not only a great impact on statistics and market insights in recent years, but it has impacted how we view important historical events as well. For those interested in looking at how the Bayesian philosophy is taking hold in many different disciplines, there is an interesting book entitled The Theory That Would Not Die.
Don’t try to prove what you already know. Let the data guide the analysis rather than allowing pre-determined beliefs to guide the analysis. Physicist Enrico Fermi pointed out that measurement is the reduction of uncertainty. Analysts start with a hypothesis and try to disprove it rather than to prove it. From there, iteration is needed to come as close to the truth as possible. The point is, an analysis that starts by trying to prove that what we believe to be true, the results are rarely surprising and the analysis is likely to add nothing new.
Think in terms of “so what.” Moving beyond the “what” (i.e., measurement) to the “so what” (i.e., insights) should be a goal of any analysis, yet many are still turning out analysis that does nothing more than state the facts. Maybe 54 percent of people in a study prefer white houses, but why does anyone care that 54 percent people prefer white houses? Analyses must move beyond findings to answer critical business questions and provide informed insights, implications and even full recommendations.
Be sure to address the “now what.” The analytics professional should make sure that the findings, implications and recommendations of the analysis are heard. This is the final step in the analytic process, the “now what” – the actual business planning and implementation decisions that are driven by the analytic insights. If those insights do not lead to decision-making or action, then the effort has no value. There are a number of things that the analyst can do to facilitate that the information is heard. A compelling story line that incorporates animation and dynamic presentation is a good start. Depending on the size of the initiative, professional videography, implementation of learning systems and change management tools should also be involved.
Just because our business technology research finds analytics as top priority and first ranked in 39 percent of organizations does not mean that adopting it will get immediate success. In order to implement a successful framework such as the one described above, organizations should build this one or a similar approach into their training programs and analytical processes. The benefits will be wide ranging including more targeted analysis, analytical depth, and analytical initiatives that have a real impact on decision making in the organization.
VP and Research Director