Prerequisites

• You have access to the Data Studio

• You had an overview of your Likelihood to Buy model

The Likelihood to Buy score is the result of the sum of all the activities of the person within the last 90 days, factored with a decay. 

Each event performed by the person is associated with an Importance (weight) and a Lifespan (decay), which are the scoring rules. MadKudu does not work with default scoring rules but instead automatically suggests custom scoring rules based on the analysis of the behavior of your past conversions. 

However, we may want to customize the weight of some events to improve the performance of the model and to suit your business need.

This means the configuration of the model essentially includes 

• configuring the importance (weight) and lifespan (decay) of each event in the Event weight tab See how to edit your Likelihood to Buy scoring model. 

• then the thresholds of points defining the very high, high, medium and low segment, in the Thresholds tab.  

In the Data Studio, the Insights tab allows to understand how events correlate to conversion and to tweak the weights and the decay of the events used in the model.

How to read the Insights graph?

The Insights graph displays the Lift of each event with regards to conversion. 

The Lift is the ratio between the conversion rate of people who did the event to the overall average conversion rate of the training dataset. The higher, the more the event is correlated to conversion. However, this depends on statistical significance. The graph is separated into 4 sections: 

• The events with the most statistical significance (at least 100 people in the training dataset have performed this event). This means there is enough data to be able to drive conclusions from the lift. The weights and decays automatically suggested are reasonable to keep.

• then the events with little statistical significance (between 10 to 100 people in the training dataset have performed this event). This means there isn't enough data to be able to drive conclusions from the lift but it still gives an indication if the events are important or not. 

• then the events with no lift available (no one in the training dataset has performed this event). This means assigning weights to these events will be taken into account in the model, but you won't be able to see their impact on the model performance.  

• then the events with no statistical significance (less than 10 people in the training dataset have performed this event). This means there isn't enough data to be able to drive conclusions from the lift as to whether or not these events are important or not. 

Within each section, events are ordered by descending lift, which lets you differentiate between the different sections: