Next Purchase Moment | Knowledge Center

What it is & the business problem it solves

Next Purchase Moment is a module of Inspire (CM's AI personalization product). In the cockpit (the front-end tooling where you adjust the business rules behind your predictions) the module is shown as "Timing Propensity", so look for that name when you configure it.

For every customer it estimates how likely they are to buy within a given number of days, and how likely they are to stop buying (churn).

Marketers usually send the same campaign to everyone, or guess who is "about to buy" or "slipping away". Next Purchase Moment replaces the guess with a daily, data-driven yes/no signal on each customer profile. Examples: "likely to buy within 30 days: yes" or "likely to churn: yes". You can use these signals to time promotions, to hold back discounts from customers who would buy anyway, and to start win-back flows earlier.

It gives a probability per time window (horizon). It does not predict an exact purchase date or an "optimal moment" to send a message.

How it works

Learn from history (weekly). Every week (Monday 03:00 by default) Inspire trains one model per horizon on past orders. The horizons are 7, 14, 21, 30, 60, 90, 180 and 365 days. The model looks at signals such as:
- how long ago the customer last bought
- how often and how much they buy
- how long they have been a customer
- their usual time between orders, and whether they are "due" compared with their own rhythm
- optionally, recent email opens and clicks, product views and add-to-carts
Score every customer (daily). Every day (05:00 by default) Inspire calculates for each customer:
- the probability of buying within each of the 8 horizons
- a churn probability: the chance of no purchase within 90 days (default churn definition)
Turn probabilities into flags. Each probability is compared with a threshold (default 0.2). At or above the threshold the flag is "true", otherwise "false". Example: a customer with a 0.35 (illustrative) chance of buying within 30 days gets "likely to buy within 30 days = true".
Deliver to CDP or CXP (daily). Three horizons are written as customer profile properties in your customer data platform (CDP or CXP). The defaults are 30, 90 and 180 days. The churn flag is always written.
- Only the true/false flags are delivered, not the probabilities.
- Delivery is incremental: only customers whose flags changed are re-sent.
- After a configuration change all customers are refreshed.
Check accuracy (weekly). Every week (Monday 06:00 by default) Inspire compares earlier predictions with what customers actually did. This accuracy check is internal and is not shown in the cockpit.

Example use cases

Scenario	Channel	Required products
A fashion retailer leaves customers with "likely to buy within 30 days = true" out of its discount campaign, which saves margin on purchases that would happen anyway	MMC Email Campaigns via a CDP or CXP segment	CDP (with Retail & Ecommerce adapter) or CXP (with Templates), Email Campaigns, Inspire Next Purchase Moment
A pet-food webshop starts a win-back workflow when "likely to churn" switches to true for a customer who bought in the last year	CDP or CXP workflow → Email Campaigns / SMS	CDP (with Retail & Ecommerce adapter) or CXP (with Templates), Email Campaigns or Campaigns, Inspire Next Purchase Moment
A DIY retailer shows a "new this season" banner only to customers likely to buy within 90 days, and a reactivation offer to the others	Web Personalization (merge tags / segments)	CDP or CXP, Webtracking, Web Personalization, Inspire Next Purchase Moment

What you can configure

Settings live on the module instance (one configured copy of the module) or on its variant (the published, versioned set of tuning settings the instance runs with). You change tuning settings in a preview (draft) and then publish it. Next Purchase Moment can run several instances and variants. Schedules are cron expressions (time expressions) in Amsterdam time.

You can configure all settings below in the cockpit. Settings marked Expert need expert rights. CM staff and customer admins can hold expert rights; request them via your CM contact person.

Setting (cockpit name)	What it does	Default	When to change	Who can change
Buy Threshold - 7 days … Buy Threshold - 365 days (8 settings)	Probability at or above which "likely to buy within N days" is true	0.2 each	Raise it for a smaller, more certain audience; lower it for reach	Standard
Churn Threshold (scoring)	Probability at or above which "likely to churn" is true	0.2	Raise it if too many customers are flagged	Standard
Churn Interval (days, training)	Days without a purchase that count as "churned"	90 days	Match your normal buying cycle (e.g. longer for furniture, shorter for groceries). Takes effect after the next weekly training	Expert
Churn Event (training)	What defines churn. Only "no purchase" is available	No purchase	Not applicable	Expert
Delivered Horizon A (days)	First horizon written to CDP or CXP	30 days	Pick the windows that match your campaign rhythm	Standard
Delivered Horizon B (days)	Second horizon written to CDP or CXP	90 days	As above	Standard
Delivered Horizon C (days)	Third horizon written to CDP or CXP	180 days	As above	Standard
Delivery Method	Write results to your customer data platform (CDP or CXP, batch) or not at all (none)	Batch	"None" for a trial without touching profiles	Expert
Event Type ID	Event type the results are written with	No default (set by CM)	At setup	Expert
Enable User Information / List of User Features	Adds customer attributes (default email and Webtracker client ID) to the delivered records so the platform can match profiles. This is an on/off switch in the Delivery section	On; email, Webtracker client ID	Rarely	Expert
Training Schedule	When models are retrained	Weekly, Monday 03:00	Rarely	Standard
Scoring Schedule	When customers are scored and delivered	Daily, 05:00	Move it before your daily campaign send	Standard
Evaluation Schedule	When accuracy is checked	Weekly, Monday 06:00	Rarely	Standard
Advanced/expert settings (CPU Request/Limit, Memory Request/Limit)	System resources	1000m CPU, 8Gi memory	Very large customer bases	Expert

The only on/off switch in this module is Enable User Information. The horizons, thresholds and schedules are always active; you tune them rather than switch them on or off.

Prerequisites & required CM products

Need	Why
CDP or CXP	Source of orders and customers, and the destination: Next Purchase Moment writes the flags to CDP or CXP customer profiles
Retail & Ecommerce adapter (CDP)	Sets up CDP's standard event types (Purchase, Add to cart, ...) that provide purchase (order line) events with order amounts. It is included in onboarding, so there is nothing extra to buy or configure
Templates (CXP)	On CXP the Retail & Ecommerce adapter does not apply. CXP uses Templates to set up the standard events instead
Webtracking	Optional. Product views and add-to-carts improve the predictions
MMC Email Campaigns / Campaigns	Optional as input (open and click events improve the predictions), and as a channel to act on the flags
Web Personalization	Optional channel for the flags
Cockpit	Configuration and statistics
Data module	Inspire's data module prepares the data for all modules (see the Data page)

Packaging tier

Next Purchase Moment is an Enterprise module: it is available in the MMC Enterprise package only. The Basic, Advanced and Pro packages do not include it. (Inspire as a whole is available from Pro upward; Pro-tier modules are also available in Enterprise.)

Data requirements

Events: purchase events with an order ID, date and order amount, linked to a known customer (with email).
Optional events: email opened and clicked, product detail view, add-to-cart. Without them the model uses order history only.
Technical minimum requirements: a customer is scored only when they have at least 1 order and their last order is within the last 365 days. This is the only minimum built into the module. There is no minimum number of customers or orders.
Cold start: known customers with no purchase in the last 365 days (including customers who never bought) get "likely to buy = false" for every horizon and "likely to churn = true". Keep this in mind when you build churn segments (see pitfalls).
More data, better flags: with very little history the model falls back to the average purchase rate, so everyone gets roughly the same probability. The more repeat buyers and history you have, the more the flags tell customers apart.

How to check it works

Customer profile properties: open a few customer profiles in CDP or CXP and check the delivered flags. For the selected horizons they are "true" or "false"; for unselected horizons they are empty. The churn flag is always filled. The delivered fields are likelyToBuy7d … likelyToBuy365d, likelyToChurn and the prediction date.
Cockpit statistics: statistics are opt-in. In the cockpit you insert a statistics template and save. The module offers the following templates, each showing how many customers are flagged true or false:
- "Timing Propensity Likely To Buy 7d Distribution" and the same for 14d, 21d, 30d, 60d, 90d, 180d and 365d (one per horizon)
- "Timing Propensity Likely To Churn Distribution"
Pick the templates for the horizons you deliver; the churn template is always relevant.
No event viewer: the module does not show per-customer results in the cockpit's event viewer.
Sanity check: the share flagged "likely to buy within 365 days" should be larger than the share for 30 days. If almost everyone is "likely to churn", check how many known customers have not bought in the last year.

Limitations & common pitfalls

Probabilities, not dates: the module answers "how likely within N days", not "on which day". It is not an "optimal send time" feature.
Never-buyers and lapsed customers count as "likely to churn": combine "likely to churn = true" with "bought in the last 365 days" (for example via an RFM label or a CDP segment) to target real at-risk customers.
Only flags are delivered: you cannot rank customers by probability in CDP or CXP. Use thresholds to make the audiences bigger or smaller.
Only three horizons at a time reach CDP or CXP (plus churn). Changing a horizon clears the old horizon's values.
Churn settings take effect after the next weekly training, not immediately.