Scenarios allow the manipulation of drivers in the dataset to see potential changes in the predicted outcome. This allows your organization to gain insight on whether increasing or decreasing key metrics, or a combination of metrics, will impact the values of the predicted outcome. This is especially useful when manipulating controllable drivers that can have an influence within the company business metrics. An example of a controllable variable might be something like digital marketing ad spend. However, manipulating drivers that can’t necessarily be controlled, such as deal-type or customer acquisition source, can provide insight to make better business decisions that impact process or resource allocation.
Scenarios work by changing the values in the dataset and then running the new dataset through the model that Kraken trained. If the variable is numeric, and the average of that value is changed, it moves that value for each row by the change in the average. So, if the average marketing spend is changed by $100, each row would add $100 to the marketing spend column and then Kraken would predict the target variable for that. If it was categorical, then each of the other columns would randomly be changed proportionate to the scale that the others are changed. For example, if the column was color and contained the colors red, yellow, and blue. The actual data shows a ⅓ split equally for each color. If the scenario changed blue to be ½, yellow and red would each randomly change some of their values to blue, leaving ½ blue, ¼ red, and ¼ yellow.