Churn Prediction

In this article, we will cover:

• Overview
• Who can use churn prediction
• How churn prediction works
  • What data it uses
  • Churn Status values
  • How often Churn Status updates
• Turning on churn prediction
• Where you can use Churn Status
  • Member Page
  • Activity and Reward Rules
  • Campaigns
  • Data Export
  • API and webhooks
• Churn prediction and RFM
• Things to keep in mind

Overview

Churn prediction is part of Advanced Analytics, an add-on to TrueLoyal Rewards. It looks at how each of your members has been buying and engaging with your loyalty program, and gives every buying member a Churn Status. This shows you who is still healthy, who is starting to drift away and who has already gone quiet.

Standard reporting tells you who has already left. Churn prediction gives you an earlier view, so you can reach members while they are still buying and there is still time to give them a reason to stay. You can then use Churn Status to target your campaigns, set up rules and send the right offer to the right members, rather than sending the same discount to everyone.

Please note that churn prediction doesn't reduce churn on its own. It tells you where to focus, and what you do with that information is down to you and your team.

Who can use churn prediction

Churn prediction is available for TrueLoyal Rewards programs where members earn points by making purchases, either online or by scanning receipts. The model needs purchase history to work with, so if your members only earn points through non-purchase activities, there won't be enough data for it to make a prediction.

How churn prediction works

What data it uses

The model looks at each member against their own buying rhythm, using the data already in your loyalty program. This includes:

• Transactional data: for example total orders, days since last purchase and total revenue.
• Loyalty data: for example available points, number of redemptions and overall engagement with your loyalty program.

Churn prediction only uses your own program's data. Your data is never shared with or used for any other brand.

Churn Status values

Every buying member is given one of the following Churn Status values:

• High Risk: the member is still buying, but their behavior looks like someone who is about to leave. This is the group you still have time to keep, so it's usually the best place to focus.
• Low Risk: the member is healthy and likely to buy again. You might want to nurture these members rather than give them extra discounts, as they are likely to come back anyway.
• Churned: the member hasn't made a purchase in over 365 days. If you want to run a win-back campaign, you can combine this with the member's last purchase date to find members who left recently, as they are often easier to win back.
• Not Available: there isn't enough activity for this member yet to make a prediction.

How often Churn Status updates

Once churn prediction has been turned on, Churn Status appears on your members within 24 hours. After that, it refreshes automatically every Saturday, so you'll always be working from the latest week's data.

Turning on churn prediction

Churn prediction isn't something you switch on yourself in the Admin Console. If you would like to add Advanced Analytics to your program, please reach out to your account team and they will take you through it. Once it is set up for your program, our team runs the first prediction for you and Churn Status starts showing on your members within 24 hours.

Churn Status becomes more reliable as the model sees more of your members' purchases, redemptions and program activity, so if your program is new, you might find the early results are less settled than they will be later on.

Where you can use Churn Status

Once churn prediction is available for your program, you can find Churn Status in several places in the Admin Console. Let's take a look at each one.

Member Page

You can see each member's Churn Status on their member profile, and you can filter the member list by Churn Status (for example High Risk, Low Risk or Churned). This lets you look at a group of members together, check their transaction history and see how they have been engaging with your loyalty program before you decide how to approach them.

Activity and Reward Rules

You can create rules that use a member's Churn Status. For example, you could award 5X points to members with a High Risk status and 2X points to members with a Low Risk status to encourage them to engage.

Campaigns

You can use Churn Status to choose the audience for a campaign, and as a condition in event-based campaigns. This means you can send a targeted offer to your High Risk members, a reminder with no discount to your Low Risk members, or nothing at all to members who are already buying regularly. That decision is down to you and what works best for your program.

If you would like to see whether a campaign actually changed your members' behavior, you can hold back a small group of members who don't receive the offer and compare the two groups afterwards. Please note that this needs to be set up manually, as churn prediction doesn't create test or holdout groups for you.

Data Export

Churn Status is included in your data exports, so you can bring it into your own reporting or BI tools.

API and webhooks

Churn Status is available in the following API responses:

• Create Member API
• List All Members API
• Update Member API with Member ID and TrueLoyal-generated ID
• Get Member API with Member ID and TrueLoyal-generated ID

It is also included in the following webhook events:

• Member Enroll Event
• Member Update Event

Based on the churn_status you receive, you can personalize what a member sees on your own site, for example by showing targeted incentives, exclusive promotions or personalized recommendations to your High Risk members.

Churn prediction and RFM

So how is this different from RFM? RFM looks at three things a member has already done: how recently they bought, how often they buy and how much they spend. It flags a member once they have already crossed a line, for example once they haven't bought for a set number of days.

Churn prediction looks ahead instead, and it also takes your loyalty data into account, such as points, redemptions and program activity. This means a member can look perfectly fine in RFM and still be starting to drift.

You can use both together. However, we recommend that you don't run an RFM win-back campaign and a churn prediction win-back campaign on the same members at the same time. If the two overlap, lean on churn prediction, as it is working from more information about the member.

Things to keep in mind

• Churn Status can't currently be used in Workflow Builder or Streaks.
• There isn't a churn prediction chart in your dashboards yet. You can find Churn Status on members, in filters, rules and campaigns, in Data Export and through the API.
• If churn prediction is turned off, any rules that use Churn Status will keep each member's last Churn Status.
• Churn prediction shows you who to focus on, but it doesn't send campaigns or create offers. You choose who to target and what to send.

If you have any questions or concerns regarding churn prediction, please don't hesitate to contact Technical Support at support@trueloyal.com.