client login
Home   /   Predictive Forecasting Becomes an Achievable Reality

Predictive Forecasting Becomes an Achievable Reality

pic_profitability

Article by Paul Davis

“Without data, you’re just another person with an opinion.” – W. Edwards Deming

As an EPM consulting leader for nearly two decades, I’ve watched with keen interest the growing phenomenon of Predictive Forecasting. To clarify, the term Predictive Forecasting is used to describe the process of generating future expected results based on the predictive analytics using data science algorithms.

Predictive Forecasting differs from normal / legacy budgeting and forecasting processes in that while typical planning processes have historically involved a very smart (hopefully) someone projecting the next several months of expected results based either on (in diminishing authority) seasonal averages, trends, gut feel, or simply compliance to an externally derived growth (hopefully) percentage target. As a result, forecast accuracy of traditional budgets are only as good as that opinion, and often suffer from poor results. Subsequently, organizations spend (hopefully?) countless hours analyzing variance of actual to target and postulating root cause for these variances. These exercises in variance analytics and root cause determination are necessary for companies to address the constant moving conditions of our ever quickening paced world.

In contrast, a Predictive Forecast removes the inherent human weightings, relying instead on actual history and a set of defined variables to feed predictive algorithms. Granted, these are a human defined set of variables, but even the factors can be improved by using data science, machine learning, and coefficient factoring algorithms to more effectively understand factor relationships. And because Predictive Forecasting is done in technical systems with the requisite horsepower to crunch Big Data volumes, deeply granular values can be used in creating a Predictive Forecast, and also be juxtaposed with actuals for variance analytics to more quickly and concretely define where forecasts are falling short.

So how is Predictive Forecasting now becoming a reality? Data Science algorithms have been around for decades. Software solutions have provided detailed analytics for years. Data Scientists themselves are able to access thousands of open source code streams in “R” to systematize their efforts. Heck, even Microsoft Excel incorporates some basic analytic algorithms.

My personal excitement about the current state of Predictive Forecasting is based on the burgeoning confluence of Predictive Analytics, Enterprise Performance Management, Analytics and Visualizations with Big Data capable database engines. By layering on the functional solutions listed above to homogenous data in columnar models, we are reaching the point of combining the functions of each business-based solution into an integrated whole;

  • Use Predictive algorithms against your Big Data store to create an informed forecast
  • Leverage visualizations and storyboards from Analytics to present variance reporting in an intuitive way that provides quick assessment of success
  • Deliver drill-back to root cause analysis in the Big Data system to evaluation transactional details

These current capabilities of Predictive Forecasting represent the next evolutionary stage of analytics:

  • Spreadsheet: Opinion / Gut Feel, fully flexible to individually create whatever desires, domain expertise
  • ERP tools: transactional, large-volume data engines built to run the business
  • Business Intelligence tools: OLAP, slice-and-dice self-service reporting, dashboards, visualizations
  • Predictive tools: algorithms, modeling, correlations
  • Embedded Platform: real-time results, proactive alerts, machine learning

Moreover, evolving Predictive Forecasting capabilities to a common platform with EPM, BI, and Big Data engines will extend the 4 V’s of Big Data to a 5th and 6th:

  1. Volume (granularity breadth and depth of data)
  2. Velocity (speed or response or computation)
  3. Variety (financial and non-financial sources, including SMAC data)
  4. Veracity (common, trusted repository of cleansed data)
  5. Visualization (interactive dashboards, charts, and the boardroom of the future)
  6. Value (alerts and actionable analysis that allows users to impact real organizational decisions)

Over the next few days, I will be posting to our SAP BusinessObjects Cloud Center of Excellence website (boc-coe.com) another couple articles with some specifics on Predictive Forecasting Use Cases, as well as how to set up and use SAP Predictive Analytics to execute two specific use cases; an algorithm for Time Series Forecast, and another algorithm for identifying influencers. I will make sure to hit on how Guided Machine Learning can be applied to these algorithms to embed them in your platform and see continual updates to result sets for improved output.

Thanks for reading.

Leave a Reply

Your email address will not be published. Required fields are marked *