The RFM (recency, frequency, monetary value) notebook is a marketing model for the BlueConic AI Workbench. You can use the model to analyze customer order data and score customers on three dimensions: Recency (how recent was the last purchase), Frequency (how frequent are purchases made), and Monetary value (how much was spent).
With RFM scores stored in customer profiles, you can create segments like "Frequent Buyers," "Recent Customers," or "Top Spenders in the Month before Holidays". This will allow you to market to specific types of customers based on their purchasing behavior. Note that in order to use the RFM notebook, you first need to have customer order data in BlueConic.
Scoring RFM (Recency, Frequency, Monetary value) with AI Workbench
The RFM notebook examines orders on the event timeline of customer profiles in a segment within a specific time period. By default, the notebook considers up to 500 order timeline events with a "revenue" field per customer lifetime, but you can configure a specific timeframe. Per customer profile, this notebook determines the most recent order, the total number of orders, and the total amount spent on all orders and stores these values on the profile. Next, the notebook scores customer profiles by bucketing these values in five equally sized score buckets (quintiles). Scores are values from 1 through 5, with 1 meaning the bottom 20% of the segment, and 5 meaning the top 20% of the segment.
For example:
A customer with an RFM Recency Score of 1 is in the bottom 20% of the segment when it comes to the most recent order. In other words: the most recent order is more towards the start date when compared to most recent orders of other customers in the segment.
A customer with an RFM Frequency Score of 3 is in the middle 20% of the segment when it comes to the total number of orders. In other words: the customer orders about as often as average customers in the segment do.
A customer with an RFM Monetary Score of 5 is in the top 20% of the segment when it comes to total amount spent of all orders. In other words: this customer is among the customers spending most in the segment.
Adding an RFM notebook
Select AI Workbench from the BlueConic navigation bar.
Click Add notebook.
A pop-up window appears. Scroll down to RFM notebook and click it.
The RFM notebook opens.
Configuring an RFM notebook
If you have the Notebook editor permissions, you can go to the Notebook editor page to initialize the parameters and to see the description of the notebook.
Go to the Parameters page.
Configure the parameters:
Select the RFM segment that defines the customers whose orders will be used for the analysis.
Select the profile property that stores the most recent order date/time.
Optional: Enter a specific number of months to look back at customer order event data.
Optional: Enter the number of days to include before the last execution.
Verify that the profile properties where the RFM values and scores will be stored are the ones you prefer and make changes as necessary.
Optional: If you'd like to add the average of the recency, frequency, and monetary value scores on the profile, select a profile property for storage.
Enter a name for the notebook and click Save to save the connection.
Running the RFM notebook
Go to the Schedule and run history page.
In the metadata section at the top of the page, you can request email notifications each time the notebook runs or only for failed runs. For details, see: setting up email notifications for AI Workbench.
Click Run now to run the RFM analysis manually.
To schedule the import and export for a future date, Enable scheduling by clicking the Settings icon. Select how to schedule the import by choosing an option from the drop-down list:
Every X minutes
Number of times per day
Days of the week
Days of the month
Weekday of the month
Set a time for the import. Click OK.
After running the notebook, you can view its output by clicking Preview. Scroll to the bottom to see a graphical representation of the RFM analysis.
Visualizing RFM notebook results with insights
To display the graph of the most recent run in your RFM notebook, add a Notebook - Single cell insight to a BlueConic Dashboard. Once added to a dashboard, configure the insight by selecting your RFM notebook and entering "16" for the cell number. This insight offers a deeper analysis of profile distribution across RFM scores, complementing the main output of the notebook, which assigns scores to each profile.
Interpreting the insight:
Bubble colors: The colors of the bubbles represent the sum of the Recency, Frequency, and Monetary (RFM) scores, ranging from 3 to 15. Warmer colors (e.g., in the (1,1,1) position) indicate lower scores, while cooler colors (e.g., in the (5,5,5) position) reflect higher scores, helping to distinguish positions in the 3D space.
Bubble size: Each bubble represents a group of profiles with a unique RFM score combination, and the size of the bubble corresponds to the number of profiles within that combination.
Score Assignment: Each consumer is assigned a score between 1 and 5 for each RFM dimension. The graph shows that the majority of profiles fall into the (Recency = 1, Frequency = 1, Monetary = 1) category, with a significant number of profiles having low RFM scores overall.
FAQs
Q: Does RFM model work well with ‘high-cost, low-frequency’ products? How low frequency is low frequency? 1x purchase a year?
A: The RFM marketing model tries to determine a certain purchase cadence. In the case of low-frequency product purchases, the BlueConic RFM model can use weeks or months instead of days to calculate the summary data.
One thing to keep in mind with low-frequency products is that you should ensure that the chosen observation period is large enough to observe multiple purchases in order to deliver the most accurate calculation.
As for how low is low-frequency, it's going to depend on how your business defines low frequency. The great thing about AI Workbench is that you can adjust the models to fit your business needs. Keep in mind that machine learning needs to have a robust set of data to work with in order to come up with conclusions. So, if you are trying to do an RFM model with limited data because you only have transaction data for a single purchase a year and a small customer base, you may need to choose a different type of model for your use case. For instance, you might want to find a model that will look at common attributes of customers to predict the buying behavior of prospects.
Q: Can I use time range for the RFM model responsively as well as explicitly?
A: AI Workbench comes with a prebuilt RFM notebook.
The RFM example notebook retrieves all orders for the last year. Marketing teams can set this time range as a parameter of the model using the AI Workbench UI. The notebook also calculates recency as the maximum of "10 – the number of months that have passed since the customer last purchased."
A marketer could configure those 10 months as another parameter. Different companies will benefit from a completely different recency formula based on the cadence of purchase or conversion events in their business. With the BlueConic AI Workbench, it only takes a few lines of Python code (the current calculation use 3 lines of code) to completely change the way the value is calculated in a way that makes more sense for your organization.