RFM Scorer

What it is & the business problem it solves

The RFM Scorer 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) it is called "RFM-Scorer". It places every customer in a simple, well-known value segment based on their order history. RFM stands for:

• Recency: how recently the customer bought

• Frequency: how often they bought

• Monetary: how much they spend per order on average

Each dimension gets a score from 1 (low) to 3 (high). The combination becomes one readable label: Champion, Loyal, Recent, Needs Attention, At Risk or Inactive. The label is written to the customer profile in CDP or CXP (CM's customer data platforms).

This lets marketers treat their best customers differently from lapsing ones without building complex rules. Examples: VIP treatment for Champions, a nudge for Needs Attention, a win-back offer for At Risk.

RFM is descriptive: it shows where a customer stands today based on past orders. It is not a churn prediction. For a predicted likelihood to buy or to churn, see Next Purchase Moment.

How it works

1. Recency score. Days since the customer's last order, compared with two boundaries. At or below the first boundary gives 3; at or below the second gives 2; above it gives 1.

2. Frequency score. Number of orders within the Recent Activity Window Size (default 365 days). The window counts back from the customer's own last order. More orders than the first boundary gives at least 2; more than the second boundary gives 3.

3. Monetary score. Average order value. Above the first boundary gives at least 2; above the second gives 3.

4. Label. The three scores together are mapped to one of the six labels. Example (illustrative): a recent, frequent, high-spending customer becomes Champion. A customer who bought often in the past but not recently becomes At Risk or Needs Attention, depending on spend.

5. Delivery. Scores and label are written to the customer profile. Only customers whose result changed are re-sent.

Automatic vs manual boundaries

For each dimension you choose how the boundaries are set.

• Manual: you enter the boundaries yourself. Example (illustrative): recency 3 at 30 days or less, 2 at 120 days or less.

• Automatic: Inspire derives the boundaries from your own customers. The percentages it uses are fixed and cannot be changed:

  • Recency: based on the typical time between a customer's last two orders. The first boundary is twice the median gap (the gap half of your repeat buyers stay under). The second boundary is the gap that 95% of repeat buyers stay under.

  • Frequency: above the median order count (more orders than half of your customers) gives 2; above the order count that only the top 5% exceed gives 3.

  • Monetary: average order values are split into roughly three equal groups (the lowest third, the middle third and the top third, at 33% and 66%).

• Automatic boundaries are calculated at the first run and then kept. They are only recalculated when the configuration changes. In a young shop with little history the first boundaries can be far too tight. Use manual boundaries in that case.

No order amount?

If orders have no amount, the monetary score stays empty. Labels are then based on recency and frequency only (RF labelling), using the same six labels.

Example use cases

What you can configure

You can configure the 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.

The scoring settings live on the variant (the published, versioned set of tuning settings a module instance runs with). You change them in a preview (draft) and then publish it. Changes made at module-instance level do not override the published variant. Schedules are cron expressions (time expressions) in Amsterdam time.

The CDP tenant and access token are not module settings. They are part of the shared profile connection that CM sets up.

Switching parts on and off: Enable User Information is an on/off switch in the Delivery section. Enable User Inclusion and Enable Item Inclusion switch their own sections (User Inclusion Settings, Item Inclusion Settings) on or off; both are off by default. If you do not see a switch for one of these two sections in your cockpit, ask your CM contact to enable it.

Prerequisites & required CM products

Packaging tier

Included from: Pro. RFM Scorer is available in the MMC Pro and Enterprise packages. The Basic and Advanced packages do not include Inspire.

Data requirements

• Orders: order ID, order date and customer, with the order amount optional (without it labels are recency/frequency only).

• Customers: customer ID and email.

• Coded rules for automatic boundaries: the run stops with an error when:

  • no customer has at least 2 orders (recency needs a gap between orders)

  • no order falls inside the activity window (frequency)

  • no order has an amount (monetary)

  Use manual boundaries until enough repeat buyers exist.

• Technical minimum requirements: with automatic boundaries, at least one customer needs 2 or more orders, at least one order must fall inside the activity window, and at least one order needs an amount (see the rules above). With manual boundaries these checks do not apply. Beyond this the module defines no minimum number of customers or orders.

• Cold start: customers without orders are not scored.

How to check it works

• Event viewer "RFM Scorer": per customer it shows the label, the three scores, days since last order, number of orders, the activity window, average order value and the boundaries used.

• Cockpit statistics: statistics are opt-in. In the cockpit you insert a statistics template and save. The module offers four distributions, each showing the number and share of customers per value:

  • RFM Segment Distribution: the share of your customers per label (for example the percentage labelled Loyal, Champion or At Risk). This is the best overview of your customer base.

  • Recency Score Distribution, Frequency Score Distribution and Monetary Score Distribution: how many customers have a score of 1, 2 or 3 per dimension.

• CDP / CXP profile: check that recency, frequency and monetary scores and the RFM label appear on a few customer profiles.

• Sanity check: if many customers who bought recently are labelled At Risk, the automatic boundaries are probably too tight. Switch to manual.

Limitations & common pitfalls

• Descriptive, not predictive: RFM does not predict churn, cancellations or lookalike audiences. Use At Risk and Inactive for win-back. For predicted churn see Timing Propensity .

• Automatic boundaries are frozen after the first run: in a young or fast-growing shop they can be far too tight, so recent buyers look At Risk. Use manual boundaries, or change the configuration so the boundaries are recalculated.

• Automatic mode needs repeat buyers: shops with mostly one-time buyers should use manual boundaries.

• Two instances writing the same CDP property conflict: each instance only sends its own changes, so labels flip back and forth. Give each instance its own property or event type.

• Edit the variant, not the instance: published variant settings take precedence.

• Schedule: the module never runs until a schedule is set.