HOW TO IMPROVE FORECAST ACCURACY IN YOUR COMPANY

"As little as one hour of training improved forecasting accuracy

by about 14% over the course of a year."

That, at least, is the conclusion Philip Tetlock and Barbara Mellers derived from the Good Judgment Project, which ran for five years (from 2011 to 2015) and recruited 25,000 forecasters to compete with four other academic research teams working on prediction issues.

In other words, even a small investment in education and training, starting with the basics, can pay off. Tetlock and his colleagues have distilled their learnings about what makes a "superforecaster," and they have written extensively about applying those ideas in a business forecasting context (for example, "Superforecasting: How to Upgrade Your Company's Judgment" and "Superforecasting: The Art and Science of Prediction").

alphachain has taken these results as an opportunity to design an exclusive series of webinars, available nowhere else on the German market, which is designed to do just that.

Starting with basic statistical forecasting methods up to state-of-the-art machine learning techniques, we explain step by step with a variety of practical examples and using state-of-the-art tools and methods (based on Excel and SAP's IBP platform) which forecasting methods offer the best cost-benefit ratio in which situation and how you can use them to significantly improve the forecasting quality in your company.

We examine some groundbreaking work in this area, presenting nine basic principles that we believe have proven their worth. 

Inspired by these principles and with a deep appreciation for the importance of tying theory to practice, we have put together the following forecasting course as a series of seven webinars in total, each of which can be taken separately or together in two blocks:

• The forecasting workflow
• The basic patterns of demand
• Statistical tools to discern demand patterns
• The basics in practice / exercises for daily use
• • Generating some patterns in Excel
  • Loading it up into IBP
  • Visualizing them in the planning view and dashboards
  • Analyzing the time series and reviewing the output

• The bread-and-butter method: Simple Exponential Smoothing
• • Generating a forecast by hand
• The mechanics of the smoothing parameter
• Calculating accuracy, bias, and stability metrics
• Forecasting a seasonal demand pattern
• Dealing with pesky outliers
• • On false positives
  • Techniques to identify and cleanse
• Simple forecasting methods in practice / exercises for daily use
• • Creating a forecasting profile
  • Generating a forecast and analyzing the results
  • Detecting and correcting outliers

• Autoregressive methods
• • The fundamentals
  • Non-seasonal ARIMA models
  • Seasonal ARIMA models

• Linear regression
• Multiple linear regression
• More advanced forecasting methods in practice / exercises for daily use
• • Parameterizing and interpreting the results of an ARIMA model
  • Parameterizing and interpreting the results of a SARIMA model
  • Parameterizing and interpreting the results of a multiple linear regression model

Foundational principles:

• The machine learning problem
• Gradient descent from scratch
• The backpropagation algorithm

• Learning the parameters of a regression model
• Classification using decision trees
• • Bagging, boosting, and growing a forest
• Classification problem - 2 examples: 1) predicting the survivors on the Titanic; 2) spam or ham?

• Machine learning forecasting methods in practice / exercises for daily use
• • Long-/medium-term forecasting problem: Gradient boosting trees
  • Short-term forecasting: Demand sensing

As we highlighted at the outset, even a few hours of training - of course, founded on sound theory - can significantly enhance one's forecasting performance within the organization (be it designing a forecasting process or using one).

Depending on the experience and interest level of the participants, webinars 1 to 4 and 5 to 7 can be booked as two separate packages independently or all webinars together for a discounted price. Details can be found on our registration form "Forecasting Course Registration" .

To get an impression of the content and approach in the webinars, we strongly recommend our free introductory webinar, for which you can also register via our website.