Churn scores in Redshift

Why use predictions for churn scores?

Knowing which of your customers is most likely to churn gives you the opportunity to act before they make that critical decision.

The most effective way to predict likelihood to churn is with machine learning. With machine learning, you can constantly keep your list of customers up-to-date with churn scores based on the historical data of similar shoppers, and plug your likely-to-churn customers right back into your stack, no PhD required. With churn scores in your stack, you're primed to jump in, offer a discount, a helping hand, or some other offer to keep them on board.

Faraday makes predicting churn scores for your customers intuitive & easy, and delivering them to any channel in your stack a breeze.

With churn score predictions in Redshift, you’ll give your team valuable insight to intervene before the customer makes that critical decision, exactly where and when they need it.

Follow the steps below to get your churn scores predictions into your Redshift account.

---

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

• Organize your customer data into cohorts
• Describe predictive models for churn scores with outcomes
• Deploy churn scores predictions to Redshift using Pipelines

Getting started with churn scores in Redshift

Make sure you have a Faraday account (signup is free!) and that it's not in test mode.

Requirements for this churn scores recipe

You'll need the following cohorts available in your Faraday account:

• A cohort representing your customers — or create one first
• A cohort representing your churned customers — or create one first

You'll also need the following connections available in your Faraday account:

• Redshift — or create one first

Building predictions for churn scores in Redshift

Now you'll create the prediction objective(s) necessary to complete this use case with Faraday.

Describe your churn scores predictions with outcomes

Outcomes use machine learning to predict whether or not people will exhibit a certain behavior.

Creating an outcome for likelihood to churn.

Let's make an outcome for likelihood to churn.

• In the navigation sidebar, choose Outcomes.
• Click the New outcome button.
• Fill out the form:
  • For Eligibility cohort, pick the cohort that best represents your customers.
  • For Attainment cohort, pick the cohort that best represents your churned customers.
  • Leave Attrition cohort blank.
  • Skip over Trait blocking.
  • Enter a memorable name, like "Likelihood to churn".
• Click the Save outcome button.

Faraday will do some magic in the background, so you can proceed with the rest of the instructions. When your outcome is done building, you'll get an email, and you can review your outcome.

Using Pipelines to deploy predictions to your stack

Now you'll configure the pipeline that deploys your predictions to redshift.

Create your pipeline for churn scores in Redshift

• In the navigation sidebar, choose Pipelines.
• Click the New Pipeline button.
• Fill out the form:
  • For Payload, choose the following:
    • Outcome: Likelihood to churn
  • For Population to include, choose the following:
    • A cohort representing your customers
  • Enter a memorable name, like "Churn scores in Redshift".
• Click the Save pipeline button.

Your pipeline will start building in the background. You can proceed immediately with the next set of instructions.

Deploying your pipeline to Redshift

Redshift

• In the Deployment area, find the Redshift module and click Add.
• Fill out the popup:
  • Provide the specified parameters for Redshift.
  • Click Next.
  • Choose the Identified option.
• Click the Next button.
• Skip the "Advanced Settings" by clicking the Finish button.
• Click the Finish button.
• Click the Test deployment button and confirm the results meet your expectations. Faraday will finish building your pipeline in the background. When it's done, you'll get an email—return to the pipeline and click the Enable pipeline button to activate it.

How to use your churn scores predictions in Redshift

Forthcoming

🔒 It's a best practice to permanently delete any file that contains personally identifiable information (PII) after use. Any deployment from Faraday that is unhashed contains PII, and should be deleted after uploading it to your destination for security purposes.