Likelihood to churn

Likelihood to churn predictions reveal which of your customers are most likely to churn, enabling you to engage those with high risk before they take that critical step.


In this tutorial, we'll show you how to:

  • Create a predictive model for likelihood to churn using an outcome.

Let's dive in.

  1. You'll need a Faraday account — signup is free!

Confirm your data


Unless you’ve already created them for another quickstart or purpose, you’ll need to add the following cohorts to your account:

  • Churned customers
  • Customers

What’s a cohort?

A cohort is Faraday’s term for a commercially significant group of people — for example, a brand’s customers, leads, or even “people who bought X and Y and then cancelled.”

Cohort membership is fluid — continuously computed by Faraday — and is defined by events its members must all have experienced and/or traits its members must all share.

For example, a Customers cohort could be defined as the group of people who have all experienced a Transaction event at least once.

For more, see our docs on Cohorts, Events, Traits, and Datasets (which define how events and traits emerge from your data).


To verify, use a GET /cohorts request. Your response should look like this:

  "name": "Churned customers",
, ...}{
  "name": "Customers",
, ...}]

Make note of the IDs of the necessary cohorts.

If the required cohorts aren’t there, follow the instructions using these buttons, then return here to resume.

Configure your prediction

Create a likelihood to churn outcome


Use a POST /outcomes request:

curl --json '{
  "name": "Likelihood to churn",
  "eligible": "$CUSTOMERS_COHORT_ID"

Your outcome will start building in the background. You can proceed immediately with the next set of instructions. When your outcome is done building, you’ll get an email.