BlueConic provides several prebuilt AI notebooks marketing teams can use (without writing any code) to apply the power of AI to your customer order data, for example to predict or analyze a customer's propensity to churn.
AI Workbench provides two notebooks that apply to customer churn predictions: this notebook to calculate customer churn predictions, and the Analyze propensity to churn notebook to analyze customer churn predictions based on customer segments. Both notebooks assume you already have customer contract or subscription start and end dates data available in separate BlueConic profile properties.
Using AI modeling to predict customer churn with AI Workbench
Marketing teams can use the Predict Propensity to Churn notebook in AI Workbench (without writing any code) to predict the probability that customers in a certain customer segment will churn. The notebook writes a churn score to the customer profile.
You can also use this AI model to predict the customer attrition for other profiles in a segment based on a particular profile property. For example, if you select 'gender' as a 'categorical profile property,' the notebook can create separate models for men and women.
The notebook analyzes customer data using an AI time-to-event model.
Running a customer churn prediction model in BlueConic
This article takes you step by step through the process of setting up and running this AI model against your BlueConic data:
- Create a notebook in AI Workbench.
- Select a specific segment of customers in the Parameters tab.
- Set the starting and ending dates of the customer's contract or subscription.
- Choose a profile property that will be used to build propensity to churn models for the different categories of the profile property.
- Select a profile property to save the churn prediction, which you can then use for segmentation or activation.
- Optional: You can adjust the AI model that calculates churn probability and select colors for the results display.
Adding a customer churn prediction notebook in BlueConic
- Select AI Workbench from the BlueConic navigation bar.
- Click Add notebook.
- A pop-up window appears. Scroll down to the Predict propensity to churn notebook and click it.
- The notebook opens to the Parameters window.
Note that if you have the Notebook editor permissions, the Notebook editor window is where you can view the notebook's Python code. It also provides details about how the notebook works.
Setting the churn notebook's parameters
- Enter a name for the notebook and Save the notebook.
- Select the Parameters tab in the left-hand panel of the churn prediction notebook window.
- For the Applied segment, choose a customer segment of profiles. This is the segment on which the churn prediction model will be trained and applied. For performance reasons, select the smallest segment that makes sense for your data.
- Select profile properties that contain these profiles' start date contract and end date contract. These profile properties should be of date-time format.
- (Optional) If you want the model to split the segment into groups based on a profile property value, select a categorial property. The model will develop churn models that compare profiles in the segment based on their values in the categorial property you select. For performance reasons, the number of categories is limited to 10.
- Select or create a churn probability profile property to save the notebook's output, the probability that a profile will churn.
- (Optional) You can adjust the maximum number of profiles the model will examine. This defaults to 2500 profiles. A higher number leads to a more accurate propensity to buy graph but is a more expensive BlueConic request. Use 0 if you want all the profiles from the segment.
- (Optional) You can change the propensity to churn model and color map used to display your results. See below to learn more about each of the propensity to churn models available in the notebook.
- Save your settings before running the model.
Selecting a propensity to churn model in AI Workbench
By default, the AI Workbench churn prediction notebook uses the Kaplan-Meier estimator to display your results. In the Parameters tab, you can choose from among several different models (and color schemes) to visualize your results.
- Kaplan-Meier estimator
- Weibull distribution
- Exponential distribution
- Log-normal distribution
- Log-logistic distribution
- Generalized Gamma distribution
Running the Predict propensity to churn AI notebook
- Go to the Schedule and run history page.
- In the metadata section at the top of the page, you can request email notifications each time the notebook runs or only for failed runs. For details, see: setting up email notifications for AI Workbench.
- Click Run now to run the analysis manually.
-
To schedule the import and export for a future date, activate Enable scheduling. Click the Settings icon
. Select how to schedule the import by choosing an option from the dropdown list:
- Every X minutes
- Number of times per day
- Days of the week
- Days of the month
- Weekday of the month
Set a time for the import. Click OK.
After running the notebook, you can view its output by clicking Preview. Scroll all the way to the bottom to see a graphical representation of the churn prediction. Or use the Notebook all cells insight to create a sharable dashboard with churn prediction graphics.
Viewing your results
When you run the propensity to churn notebook, a preview window opens with your results for relevant sections of the code or notebook cells. This notebook also writes the churn probability to each profile it evaluates, in the profile property you selected in step 6 above. You can use this information to target or segment profiles with offers to prevent churn.
Note that you can use the Notebook all cells insight to create a sharable dashboard with churn prediction graphics.
Companion AI notebook for analyzing customer churn propensity
You can use the companion AI notebook, Analyze propensity to churn, to examine how churn risk compares between different customer segments.