Accuracyfit of predictive models executive buy

This preview shows page 8 - 10 out of 13 pages.

Accuracy/fit of predictive modelsExecutive buy-in/approval/acceptanceBusiness case content/formulationInternal people (skills, expertise)Budget availability/sizeTechnologies/tools/infrastructureInterpretation/visualization/communication of resultsPlease rank each of the following key success factors that inhibit your predictive modeling and analytics capabilities – from high importance to low importance:0%20%40%60%80%100%High Importance/Medium-High ImportanceMedium ImportanceMedium-Low Importance/Low ImportanceData (availability, quality,governance, privacy laws)Internal people (skills, expertise)Budget availability sizeAccuracy/fit of predictive modelsExecutive buy-in/approval/acceptanceTechnology/tools/infrastructureInterpretation/visualization/communication of resultsBusiness case content/formulation65%23%12%60%27%13%60%23%17%59%26%15%58%21%21%56%26%18%56%26%18%54%27%19%
8 PROTIVITI • 2016 Protiviti Predictive Analytics SurveyAchieve Better Precision and AccuracyA predictive model can only be relied upon if it is accurate. Although we have observed a definite improvement over the past few years, model precision generally averages 80 percent, depending on the type of application, data availability and quality, and modeling technique used, among other factors. For a majority of organizations, this shortfall represents a major point of frustration.Some have become accustomed to unsystematic ways of making decisions, typically based on a “gut feel,” which is less predictive than a coin flip. However, the picture is much clearer for market leaders and executives who use predictive analytics to gain and sustain a competitive advantage. They are able to rely on models that are at least 92 to 95 percent accurate. To reach that level of success, you need more and better quality data, as well as better models – and not simply from a quantitative engine perspective. The model definition and design team needs to be comprised of cross-disciplined subject-matter experts skilled in model design and capable of thinking outside of the box. In addition, the entire model development cycle must be supported by efficient processes and governed properly. Alignment between people, processes and technology is critical to achieving the best performance and accuracy from your predictive models.Another interesting trend is the move to machine learning algorithms and artificial intelligence, which are a driving force for unstructured data and text analytics. From a tool perspective, “freeware” (like Python and R languages) are steadily gaining ground against the most commonly used commercial statistical engines and platforms. rank the accuracy and fit of predictive models to be highly important for their predictive modeling and analytics capabilities.

  • Left Quote Icon

    Student Picture

  • Left Quote Icon

    Student Picture

  • Left Quote Icon

    Student Picture