Exploring Customer Lifetime Value (CLV) model

Overview

The CLV model dashboard is a visual representation of the data and insights generated by the Skypoint AI-powered CLV model. It displays various metrics and charts that provide a comprehensive view of customer lifetime value.

The predicted value of a customer is based on the history of business transactions. With the help of metrics and graphs, you can track and analyze the financial value of your customers over the course of their relationship with a business. It helps to identify and target high-value customers, optimize marketing and sales strategies, and make informed business decisions.

View CLV model

Follow the below steps to explore the CLV model:

1. Go to Predictions > Built-In.
2. In the My predictions tab, click on your CLV model.

The CLV model page appears.

Analyze CLV model

There are four main sections of data displayed on the view page. You can analyze the data of your customers and identify trends, patterns, and insights. The CLV model helps you with the following:

• Identify high-value customers and reward them.
• Create a strategic audience to run personalized campaigns.
• Optimize sales and marketing strategy.

Model performance

Scores A, B, or C in a CLV model are used to indicate the performance of the prediction, with A indicating the best performance and C indicating the worst. These scores can help determine the accuracy and reliability of the model's predictions and aid in making decisions about whether to use the results stored in the output table.

The scores for the Skypoint CLV model are determined by comparing the model's accuracy and performance against a set of predetermined rules. These rules are as follows:

Score	Description
A	When the model accurately predicted at least 5% more high-value customers than the baseline model.
B	When the model accurately predicted between 0 to 5% more high-value customers than the baseline model.
C	When the model accurately predicted fewer high-value customers than the baseline model.

note

The baseline model uses a non-AI-based approach to calculate customer lifetime value based primarily on historical purchases made by customers. If a model receives a grade of A, it will be considered highly accurate and suitable for use in a production environment, while a model with a grade of C would be considered unreliable and may need to be retrained or replaced.

Performance metrics

Skypoint AI performance metrics include the following:

• MSE - Mean squared error (MSE) of an estimator measures the average of the squares of the errors. The lower the value, the better the model. The best value is 0.0.
• Accuracy - Accuracy is defined as simply the number of correctly categorized examples divided by the total number of examples. The best value is 1. The higher the value, the better the model. The disadvantage of accuracy is that it is not robust when the data is unevenly distributed, or where there is a higher cost associated with a particular type of error.
• F1 - It is the harmonic mean of precision and recall value. It can even correctly measure the model performance when there is a large class imbalance. The best value is at 1 and the worst is at 0.
• WeightedPrecision - Precision is the ratio between the true positives and all the positives. Weighted precision gives the weighted mean of precision with weights equal to class probability. A higher value indicates a higher accuracy of the positive predictions.
• WeightedRecall - The recall is the measure of our model correctly identifying true positives. Weighted recall gives the weighted mean of recall with weights equal to class probability. A higher value indicates a higher completeness of the positive predictions.
• AreaUnderROC - The Area Under the Curve (AUC) is the measure of the ability of a classifier to distinguish between classes and is used as a summary of the ROC curve. The higher the AUC, the better the performance of the model at distinguishing between the positive and negative classes. The best value is 1 and the worst value is 0.
• AreaUnderPR - The precision-recall curve is constructed by calculating and plotting the precision against the recall for a single classifier at a variety of thresholds. The higher the AUC, the better the performance of the model. The best value is 1 and the worst value is 0.
• RMSE - Root Mean Square Error (RMSE) is the standard deviation of the residuals (prediction errors). The lower the value, the better the model. The best value is 0.

Value of customers

A value of customers graph in a CLV model is a visual representation of the distribution of customer lifetime values, where the x-axis represents the different CLV ranges (e.g., high, low) and the y-axis represents the number of customers within each range. For example, if a business has a high number of customers with high lifetime values, it indicates that the business is effectively retaining and upselling to these customers.

Most influential factors

There are many factors that are considered when creating your prediction. For example, information about a customer's past purchases, such as the frequency and value of purchases, can provide insight into their future spending potential.

On the Most influential factors graph, the Explanation key column represents the factors, and the Weight column represents the importance or weight of each factor.

The length of each bar shows the importance of each feature in the model. For example, if the model finds a factor such as Overall Transaction Value is highly influential in determining customer lifetime value. The company may focus on increasing this factor through tactics such as upselling or cross-selling to customers and creating attractive offers and discounts for high-value customers.

Create a quick audience

Creating a quick audience from a CLV model involves identifying a group of customers based on their potential value to run personalized campaigns with targeted sales, marketing, and support efforts. For example, customers who have a high CLV can be segmented into a group of high-value customers and targeted with personalized offers and incentives to increase their lifetime value. On the other hand, customers with a low CLV can be segmented into a group of low-value customers and targeted with retention campaigns or other strategies to increase their lifetime value.

Follow the below steps to create an audience from the churn model:

1. On the churn model view page, click Create Audience.

2. Choose the Customer lifetime value such as High or Low from the dropdown list. After selecting the CLV, you can view the number of profiles for the selected CLV out of the total profiles.
3. Enter the Name and Output table name to identify your audience.
4. Click Create.

Once you have created the audience, you can use it to target your marketing campaigns or to create personalized recommendations for them. You can view the new audience under Activate > Audiences.