Case Study
Tune the Market Simulation – Part 02 – Product Tuning
The previous workflow CS-123 Cola Market 2015 Feature Tuning only partially tuned the Market Simulation for Cola. The Input Parameters that defined the differentiation of individual Features were tuned to reduce the overall ‘Quantity Error’. That is, the error between the Market Simulation predictions and the real-world observations were minimized. But Markets cannot always be entirely explained by the part-worth values that Customers place upon individual Features.
Product Tuning makes it possible to further tune a Market Simulation so that it is even more accurate. Product Tuning can be done with a ‘Tune Market’ node (available as part of the Scientific Strategy Premium Edition).
Product Tuning works on the final Willingness To Pay (WTP) Matrix generated by the Feature Tuning workflow. Details of the individual Features are lost as Product Tuning works to adjust only the WTP Customer Distribution of each Product.
This Case Study provides a high-level overview of the workflow without detailed explanation. It assumes you are already somewhat familiar with KNIME and Market Simulation. If not, start by reviewing the Building Blocks and Community Nodes.
Downloads
Product Tuning
#1 Tune WTP Matrix
Tune Market
The Input Product Array (top-port) contains the real-world Prices and Quantities for each Product.
The Input WTP Matrix (bottom-port) contains the final results from the previous Feature Tuning workflow.
After about 2 minutes of calculations and 141 internal iterations, the Sum of Total Error is reduced down to 420 Customers. That is, of the purchase decisions of nearly 100,000 real-world Customers, only 420 (0.4%) cannot be explained by the Market Simulation.
Click on an icon below to see and scroll through the enlarged version of the images.
