RFM Model

Create an RFM model from imported order summary or order detail data to segment and analyze users by purchase recency, frequency, and monetary value.

What is an RFM model?

The RFM model measures user value through three indicators: Recency (R), Frequency (F), and Monetary (M).

The model scores each user on all three metrics. Example scoring criteria:

R: Days since last purchase. 0–30 days: 3 points, 31–90 days: 2 points, above 90 days: 1 point.
F: Number of purchases. 1 time: 1 point, 2–4 times: 2 points, 5 or more: 3 points.
M: Total spend. 0–100: 1 point, 100–1000: 2 points, above 1000: 3 points.

Each user's score is compared against a baseline (the group average or a custom value) to determine relative value. Combining all three indicators divides users into 8 types for targeted operations.

RFM User Types and Classification Rules

By comparing each user's R, F, and M scores against their respective baselines, users are classified as high or low on each dimension:

Score ≥ baseline: High
Score below baseline: Low

Note

RS, FS, and MS represent the Recency, Frequency, and Monetary scores of each user.
The RS, FS, and MS baselines default to the weighted average of all users in the model, or can be set to custom values.

To configure scoring rules and baselines, see Create an RFM model below.

Each R, F, and M dimension is classified as high or low, producing 8 user types:

RFM customer type	RS	FS	MS	Description
High-value users	Greater than or equal to RS comparison value	Greater than or equal to FS comparison value	Greater than or equal to MS comparison value	Recent purchase, high frequency, and high spending.
Key retention users	Low	Greater than or equal to FS comparison value	Greater than or equal to MS comparison value	Last purchase is distant, but frequency and spending are high.
Key development users	Greater than or equal to RS comparison value	Low	Greater than or equal to MS comparison value	Recent purchase and high spending, but low frequency.
Key winback users	Low	Low	Greater than or equal to MS comparison value	Last purchase is distant and frequency is low, but spending is high.
General value users	Greater than or equal to RS comparison value	Greater than or equal to FS comparison value	Low	Recent purchase and high frequency, but low spending.
General retention users	Low	Greater than or equal to FS comparison value	Low	Last purchase is distant and spending is low, but frequency is high.
General development users	Greater than or equal to RS comparison value	Low	Low	Recent purchase, but low frequency and low spending.
Potential users	Low	Low	Low	Last purchase is distant, with low frequency and low spending.

Create an RFM model

Two analysis types are available:

Order summary data: aggregates raw data from the last N days into a per-user table with one row per user.

For more information about the sample Quick Audience Import Data Table Requirements, For more information about how to import order summary table data, see Order Summary.
Order detail data: uses individual transaction records as the analysis basis. The system aggregates order details per user when generating the model.

For more information about sample order details, see Quick Audience Import Data Table Requirements. For more information about how to import order details, see Order Details. For more information about how to report and store order event data, see Event Hub.

Procedure

Go to Workspace > User Insight > Marketing Model > RFM Model.
Click New in the upper-right corner to open the RFM model configuration page.
Configure the model as follows.
1. Select Order Summary or Order Details as the analysis type.
2. Select the imported data table to analyze (use the data table alias set during import).
3. Select the currency unit: CNY, USD, GBP, EUR, or HKD.
4. Click Next.
Configure RFM parameters

RFM analysis of order summary data:

RFM analysis of order details:
1. For Order Details only: specify the statistical period, order behavior type, and order time fields.
  
  If the Purchase Channel and Commodity Category Name fields were configured during import, you can also filter data by statistical channel and commodity category for targeted analysis. Each supports up to 100 values.
2. For all types: Select the number of scoring intervals (3 or 5) for R, F, and M, and set the range for each score. The user distribution per interval is displayed below.
  
  Scoring rules:
  - R (Recency): Fewer days since last purchase = higher score.
  - F (Frequency): More purchases in the period = higher score.
  - M (Monetary): Higher spending in the period = higher score.
3. Click Next after configuring the R, F, and M scoring rules.
Configure the comparison baselines.

Baselines determine each user's relative value level and user type classification. For more information, see RFM User Types and Classification Rules.

RS, FS, and MS are the scores of R, F, and M respectively. You can use the average score of all users as the baseline, or set custom values.
- If you use the average (weighted mean), the page displays the current RS, FS, and MS averages based on the configured scoring rules.
- If you set custom values, use the displayed averages as a reference and adjust based on your business needs.
Click Finish, enter a model name in the dialog box, and click Save.

The new model appears in the RFM model list. Available management operations are described in Manage RFM Model.

Manage RFM models

The following operations are available for RFM models:

Edit: Click Edit to modify the model configuration.
Analyze: Click Analyze to run RFM analysis. For more information, see RFM Analysis.
Update: Click Update to refresh the model with the latest data.
Rename: Choose > Rename to change the model name.
Delete: Choose > Delete and confirm.

Note
Models with generated audiences cannot be deleted.
Permissions: Grant non-administrators access to use or manage RFM models. Configure permissions as described in Authorize Tags.
Note
- Non-administrators see RFM models on the Authorized tab of the RFM Models page.
- By default, administrators can manage all RFM models in the workspace and see them on the My tab.
Update settings: Choose > Update Settings. Enable Auto Update and set the update schedule. The model updates automatically when the underlying order details table or order summary table completes its scheduled task within the specified date range.

Note
If update parallelism is configured in Central Administration, models exceeding the limit are queued. For more information, see Workspace System Configuration.