Performance Assessment
Accurate data collection and analysis are fundamental processes in assessing business performance. Common metrics such as sales revenue, product cost, and profit margin are the basic reporting or accounting components required to measure ‘bottom line’ success. Unfortunately, the overwhelming majority of organizations employ only ‘Descriptive Analytics’ – the organized reporting of past events – to inform future decision-making, an analytics methodology ill-suited to support the increasing demands of digital customer engagement.
Post-Mortem Limitations
Descriptive Analytics investigates past events and actions and implicitly assumes that future events and actions will follow the same approximate patterns. The variables can be simple counts or summaries of digital customer engagement parameters such as website visits, program registrations, and ecommerce conversions. Too often, basic mathematical operations that calculate elementary numerical ratios such as click-through, bounce-rate and conversion rates are classified as Descriptive Analytics. Although Descriptive Analytics are useful, the obvious flaw is that many portions of past behavior might not be indicative of future behavior, and too many marketers rely on these narrowly defined ‘key performance indicators’ to make important decisions.
Marginal Enhancements
Many marketing organizations claim to leverage ‘Advanced Analytics’. Regrettably, the Advanced Analytics tools that most Business Intelligence professionals are referencing are really no more than minor enhancements to rudimentary reporting capabilities. Data filters, sorting functions and conditional constructs simply provide alternative outcomes or perspectives on Descriptive Analytics. For example, a data filter applied to sales transactions might deliver sales per week, sales per location, or sales per category, yet each of these scenarios is a simple reformatting of existing data, not advanced analytics per se. A recent University of Cambridge Service Alliance paper on Data and Analytics estimates that nearly 80% of all business analytics today are fundamentally descriptive in nature. This statistic infers that most companies “are driving forward while looking into the rear-view mirror”, a rather precarious and unpredictable situation. In the digital engagement ecosystem, a forward-looking approach is the only viable option.
Predictable Outcomes
Predictive Analytics is the most recent evolution of business intelligence, and is rapidly gaining acceptance. It includes a variety of statistical techniques from predictive modeling, machine learning, and data mining that analyze current and historical data to make future predictions. The central elements of Predictive Analytics are the predictors, or variables, that can be correlated and measured to predict future behavior with high-levels of statistical certainty.
The promise of Predictive Analytics is very compelling and many analytics practitioners argue it is the future of all business intelligence methodologies. It allows marketers to look forward and see not only the possible outcomes, but also the likelihood of those outcomes actually occurring by changing specific variables. Predictive Analytics leverages the same data as Descriptive Analytics, and often much more, to indicate future events. Sophisticated predictive modeling tools can ingest vast quantities of big-data, model the outcome, score the results and deploy the algorithms all in real-time.
Proven statistical techniques commonly used in Predictive Analytics often include regression and/or time-series modeling to detect, measure and score different trends, patterns and relationships between both relational and transactional big-data sets.
Enterprise-wide Commitment
Building an internal Predictive Analytics capability can be an arduous time-consuming task, and many marketers opt to outsource the capability. First, it requires a broad variety of data management and analytics tools, from the relatively simple to rather complex, depending on the company’s business requirements. Second, it requires the selection, or upgrading, of both internal and external application platform(s) and tools such as SQL, R-Studio, Hadoop, Spark, Tableau, etc. Third, it requires a significant investment in data technology, management, and analytics resources so power-users can leverage the capability on an enterprise-wide level.
The ROI behind Predictive Analytics, or smarter decision-making, is directly correlated to the incremental sales or savings realized. Overall success is heavily predicated on a company culture willing to embrace a truly data-driven approach to improve the business. Empirical evidence suggests that the positive economic opportunities associated with Predictive Analytics are quite significant. Diverse industries such as insurance, healthcare, transportation, retail, media, and manufacturing – even professional services – have impactful use cases with strong tangible and intangible ROI.
RevTrax has envisioned and pursued the following predictive modeling scenarios for digital promotional incentives:
- Retailers using digital incentives to predict in-store foot traffic, category shelf movement and staffing requirements during promotional periods
- CPG manufacturers using digital incentives to predict the right product, the right discount at the right moment to trigger shopper activation for a purchase transaction.
- Marketers using digital incentives to optimize digital marketing channel attribution by predicting the success and associated cost of each program prior to implementation
Sadly, fewer than 20% of organizations have embraced Predictive Analytics as a business-building tool. RevTrax believes the barriers are two-fold. First, it is a commitment that must be embraced at all levels of the company and that requires a thorough understanding of the risks and rewards, a knowledge base that is still relatively restricted to the technical and data functions. Second, only the executive leadership can make the decision to make significant investments in highly qualified staff of data strategists, analysts, scientists, etc. required to implement an internal Predictive Analytics capability.
Analytics Frontier
Prescriptive Analytics is possibly the final phase of business intelligence methodologies that include Descriptive and Predictive Analytics. It automatically aggregates big-data, information technology, mathematical disciplines and strategic decision rules to make predictions and then suggests the decision options that will achieve the desired outcome. Prescriptive analytics is capable of incorporating a large number of business rules, ingesting new big-data and new variables in real-time and retraining existing models on-the-fly, thus improving the prediction accuracy of likely outcomes with near complete automation.
Predictive Analytics arguably anticipates the future, but more importantly, it answers the questions of when, why, where, and how a certain event will occur and suggests decision options that can impact those future outcomes. It is not only the future of business intelligence; it is the future of all data-driven intelligence regardless of the use case. Prescriptive Analytics leverages the core foundation of Descriptive Analytics and automates the fulfillment of desired outcomes made by Predictive Analytics, making it significantly easier for marketers to make the best possible decisions.
The retailer, predicting daily foot traffic by store, can now prescribe the absolute best timing of their promotions. The marketer can now engage shoppers on their quest for a new brand with the right message, and the right offers, at the right times and ultimately convert them to lifelong brand loyalists. This is the future of business intelligence.
Practical Solutions
Predictive and Descriptive Analytics, when used in tandem, foster better decision-making by providing a window into both past and future performance. Prescriptive Analytics is a step-change that automatically predicts the desired outcomes and provides the opportunity for marketers to influence those outcomes by changing the variables ahead of time. Using these mathematical modeling techniques, marketers can identify and understand the relationships between multiple different factors and act accordingly with the most informed decision making. The practical choice is that which best meets the wants and needs of the organization.
