Predictive analytics gives you an indication of how your customers will behave in the future, using artificial intelligence to get insights from the history of their purchases.
RFM analysis tells you how it went: how recently, how often and how much each contact has spent. Predictive takes the next step and tells you what is about to happen: which customers are about to stop buying, how likely they are to do so and by when.
With Churn prediction magnews estimates, for each of your customers, how likely they are to stop purchasing and by when churn is expected, and places them in a churn segment that goes from Loyal to Churned.
You turn it on for the individual RFM model, from the General tab of the Settings.
How it works
- It learns from your data, not from rules you set. The model analyzes the order history and derives the thresholds from the purchasing habits of your customers. A subscription coffee shop and a jewelry store have completely different repurchase rhythms, and the same configuration produces sensible results in both cases.
- It updates itself. The prediction is recalculated once a day, so it follows customers as they buy or drift away, without you having to review anything.
- The results become contact data. They are written to contact fields, so you can use them like any other contact data: in filters, in audiences, in workflows and in message personalizations.
- You need an order history. The predictions are based on the individual orders, so Churn prediction is available only on models that read a data table, with enough history behind them.
The typical use case is a retention campaign that starts automatically when a customer enters the At risk or Imminent churn segment.
An example. Marta has bought three times in the last eight months, the last one in September. The model recognizes that customers with that behavior buy again within sixty days on average, estimates a churn probability of 82% and places the expected churn date ten days from now: Marta enters the Imminent churn segment. That same day a workflow sends Marta a dedicated offer, without you having to check by hand who is drifting away. If they buy, at the next calculation they are back among the Loyal ones and the revenue they have generated shows up in the recovered revenue of their audience.
What you need to use it
Churn prediction is based on the history of the individual orders, so it is available only on the RFM models that read a data table, with the data source already configured. It is not available on models that use Contact Fields, where all you have is the aggregate data of last purchase date, number of purchases and total spend. If you use one of the native eCommerce integrations (Shopify, Magento, Prestashop) that feeds a data table, you are already halfway there.
You also need enough order history: Churn prediction gives the best results on a database with at least a year of orders and a significant share of customers who have purchased more than once. If your eCommerce has just started, it is better to wait until you have built up some history.
You find the complete prerequisites and all the configuration steps in Setting up Churn prediction and reading the results.
What you get
The calculation runs once a day and writes the results to five contact fields: the expected churn date, the churn probability, the churn segment, the previous churn segment and the date of last segment change. They are ordinary contact fields, so they go into filters, audiences, workflows and personalizations exactly like the others.
The same data is also available at a glance: on the contact home you find the Churn propensity card with segment, probability and expected date, while on the database dashboard and on audiences you find the distribution of the contacts across the six segments and a financial dashboard that answers the question that counts when you have to justify an investment in retention, that is how much churn is costing you and how much of it you are recovering.
Using churn data in your activities
Since the five output fields are ordinary contact fields, you can:
- filter and create audiences on the churn segment or on the churn probability, for example the audience of the customers in Imminent churn with a high spend;
- start workflows when the segment changes, using the Previous churn segment field and the Date of last segment change;
- personalize the content of your communications based on the segment;
- cross churn and RFM clusters, to tell for example a VIP at risk from a lead at risk and reserve the more generous incentive for the first one.
Frequently asked questions
Does churn replace the RFM clusters?
No, it completes them. RFM describes the current value of the customer, churn estimates the risk of losing them. Together they tell you who is worth investing in first.
Do I have to set the risk thresholds myself?
No. The model derives the most delicate values from the purchasing habits of your customers, so the default configuration works even on businesses with very different repurchase rhythms. If you know your repurchase cycle better than the model can infer it, you can still work on the advanced parameters.
How often is the prediction updated?
Once a day, based on the schedule of the RFM model. You can also start the calculation manually at any time.
Can I have more than one model with churn enabled?
Yes. Every model has its own output fields and its own scheduled activity, and a panel per model appears in the cards.