machine-learning · business

Subscription Churn Model Confusion Matrix

This confusion matrix evaluates a subscription model that predicts whether a customer will stay or churn in a defined period. The churn-to-retain cell represents missed churn risk, while retain-to-churn represents customers who might receive unnecessary retention outreach. The correct balance depends on outreach cost, customer experience and expected value, not accuracy alone. Teams should set the outcome window before calculating the matrix and use a holdout period that reflects current customer behavior. This example uses aggregate counts and is intended for model evaluation, not for making automated customer decisions without appropriate review.

Type echarts-heatmapEngine echarts-heatmapUpdated 9/24/2026

Subscription Churn: Actual vs Predicted

Drawing preview

The request

“Create a binary confusion matrix heatmap for a subscription churn model with actual rows Retain and Churn and predicted columns Retain and Churn. Use counts: actual Retain predicted Retain 640 and Churn 82; actual Churn predicted Retain 37 and Churn 141.”

Then tryRename Retain to Stay.Add a note that the evaluation covers the June cohort.

What is in this drawing

Read the decisions behind it.

01Define the outcome window

State when churn is measured.

02Review missed churn

These are actual churners predicted to stay.

03Review unnecessary outreach

Predicted churners who stay may receive offers.

Use for prioritization analysis with privacy-reviewed, aggregated results.

More like this

Other examples in this family.

Drawing preview

Email Intent Classifier Confusion Matrixmachine-learningOpen →

Drawing preview

Helpdesk Ticket Routing Confusion Matrixmachine-learningOpen →

Drawing preview

Invoice Fraud Screening Confusion Matrixmachine-learningOpen →

Browse all heatmap templates →

FAQ

Common questions

What is missed churn?

An actual churner predicted to Retain.

Why does false churn matter?

It can trigger unnecessary outreach or incentives.

What is the outcome window?

The defined period used to decide whether a subscriber churned.