Node Description

Similarity Matrix Node

The Similarity Matrix node is designed to take a super-set of Products and determine which of those Products should be included in the Market based upon their Similarity Rankings and Quantity sold.

Starting with the user-defined Focus Product, the Similarity Matrix node determines the distance to each of the other Products based upon the Input Product Similarity Rankings. The distance reflects the ‘Difference Probability’ – that is, the probability that the Focus Product and the other Product are different. If the ‘Difference Probability’ is very small (say, 0.01) then customers view the Focus Product and the other Product as nearly identical – the Products are highly correlated and the distance between them is small. On the other hand, if the ‘Difference Probability’ is very large (say, 0.99) then customers view the Focus Product and the other Product as uncorrelated. Note that Correlation equals one minus ‘Distance Probability’.

After distance from the Focus Product has been calculated the Similarity Matrix node selects the Products to include in the final Market and calculates the Output Product Correlation Matrix. Product selection depends upon Products that have a high degree of Correlation. But Product selection can also take into account the Quantity sold of a Product. If the Quantity sold of a Product is large then the Product may be included even if its distance from the Focus Product is also large.

Finally the Similarity Matrix will calculate an Output Brand Correlation Matrix. The Brand Correlation Matrix includes all Brands found in the Input Product Array. The Brand distance calculation is similar to the Product distance calculation, but the relative number of Brands to Products is taken into account. If there are a lot of Products but only a few Brands then the ‘Difference Probability’ is increased by the ratio of Products to Brands. The final distance will then naturally be reduced back down by the number of Brand matches in the Input Product Similarity Rankings table.

This Premium Node is not available as part of the free Community Edition. Premium Nodes help clean and connect real-world data to Market Simulations, and provide advanced Market Science analysis. Note that these descriptions are often deliberately vague.

Downloads

Similarity Matrix

The Similarity Matrix node is designed to take a super-set of Products and determine which of those Products should be included in the Market based upon their Similarity Rankings and Quantity sold. Only those Competitive Products most relevant to the ‘Focus Product’ are included in the Market.

Inputs

Input Product Array

The super-set of Products from which the Market is defined. Each row corresponds to a Product that is to be compared with the Focus Product.

Product Similarity Rankings

The list of all Product-to-Product rankings. That is, from the Customers who looked at Product01 how did they rank Product02. Alternatively, from the Customers who looked at Product01 what is the probability that they also looked at Product02 (sorted to provide a rank order).

Node

Configuration

The user sets the probability that the Product with the top ranking of 1 (when compared to the Focus Product) is correlated with the Focus Product. The Correlation Probability must be less than (not equal to) 1.0. A high Correlation Probability of between 0.9 and 0.999 is typical.

Outputs

Output Product Array

The Output Product Array corresponds to the Input Product Array but updated to reflect information about selection in the Market and about each Product’s correlation with the Focus Product.¬†A boolean ‘Include’ value indicates whether or not the Product was selected for inclusion in the Market and the Output Product Correlation Matrix.

Product Correlation Matrix

A symmetrical matrix reflecting the correlation between each Product and each of the other Products included in the Market.

Brand Correlation Matrix

A symmetrical matrix reflecting the correlation between each Brand and each of the other Brands in the Market.