Approximate RFM

RFM is a method used to analyze customer value that represents three dimensions:

  • Recency. How recently did the customer purchase?

  • Frequency. How often does the customer purchase?

  • Monetary. How much does the customer spend?

A score for recency, frequency, and monetary is assigned on a scale of 1-10, and is then aggregated into a combined RFM score that is assigned on a scale of “111” (the lowest possible RFM score) to “101010” (the highest possible RFM score, or “three ‘10’ scores”). Maximum scores represent preferred behaviors.

What is an RFM score?

An RFM score is an approximation that measures three data points:

  1. A recency score (R) that sorts customers by how recently they purchased during the previous 12 months.

  2. A frequency score (F) that sorts customers by purchase frequency during the previous 12 months.

  3. A monetary score (M) that score that sorts customers by spend amount during the previous 12 months.

These individual scores can be combined into a single score (RFM).

Use approximate RFM (recency, frequency, monetary, and combined score) attributes to build segments that support many types of use cases.

Approximate RFM attributes include:

Examples

The following topics contain examples of using approximate RFM scores:

Recency, frequency, and monetary

You can access individual RFM scores directly from the Segment Editor. To add any of these attributes to your segment, click Add criteria, select the Transaction Attributes Extended table, select L12M Frequency, L12M Monetary, or L12M Recency, and then apply an condition.

L12M Recency

L12M Recency is a score that sorts customers by how recently they purchased during the previous 12 months.

Each RFM score is split into ten percentile groups. The lowest percentile is 1 and the highest percentile is 10. Each percentile represents 10% of the customers who belong to that segment.

  • 10 represents the 90-100th percentile and the customers who have the highest recency, frequency, or monetary scores.

  • 9 represents the 80-90th percentile

  • 8 represents the 70-80th percentile

  • 7 represents the 60-70th percentile

  • 6 represents the 50-60th percentile

  • 5 represents the 40-50th percentile

  • 4 represents the 30-40th percentile

  • 3 represents the 20-30th percentile

  • 2 represents the 10-20th percentile

  • 1 represents the 0-10th percentile and the customers who have the lowest recency, frequency, or monetary scores.

Tip

Combine percentiles to build larger groups of customers. For example 9 and 10 together represent the “top 20%” while 8, 9, and 10 represent the “top 30%”.

To find customers who purchased most recently, start with the L12M Recency attribute in the Transaction Attributes Extended table, set its condition to is, and then specify a percentile. For example, use “10” to find customers who are in the top 10% for most recent purchases.

Choose the L12M recency attribute from the Segment Editor.

The attribute will appear in your segment like this:

Find approximate RFM recency.

L12M Frequency

L12M Frequency is a score that sorts customers by purchase frequency during the previous 12 months.

Each RFM score is split into ten percentile groups. The lowest percentile is 1 and the highest percentile is 10. Each percentile represents 10% of the customers who belong to that segment.

  • 10 represents the 90-100th percentile and the customers who have the highest recency, frequency, or monetary scores.

  • 9 represents the 80-90th percentile

  • 8 represents the 70-80th percentile

  • 7 represents the 60-70th percentile

  • 6 represents the 50-60th percentile

  • 5 represents the 40-50th percentile

  • 4 represents the 30-40th percentile

  • 3 represents the 20-30th percentile

  • 2 represents the 10-20th percentile

  • 1 represents the 0-10th percentile and the customers who have the lowest recency, frequency, or monetary scores.

Tip

Combine percentiles to build larger groups of customers. For example 9 and 10 together represent the “top 20%” while 8, 9, and 10 represent the “top 30%”.

To find which customers purchased most frequently during the previous 12 months, start with the L12M Frequency attribute in the Transaction Attributes Extended table, set its condition to is in list, and then specify percentiles. For example, use “10” and “9” to find customers who are in the top 20% for purchase frequency.

Choose the L12M frequency attribute from the Segment Editor.

The attribute will appear in your segment like this:

Find approximate RFM frequency.

L12M Monetary

L12M Monetary is a score that sorts customers by spend amount during the previous 12 months.

Each RFM score is split into ten percentile groups. The lowest percentile is 1 and the highest percentile is 10. Each percentile represents 10% of the customers who belong to that segment.

  • 10 represents the 90-100th percentile and the customers who have the highest recency, frequency, or monetary scores.

  • 9 represents the 80-90th percentile

  • 8 represents the 70-80th percentile

  • 7 represents the 60-70th percentile

  • 6 represents the 50-60th percentile

  • 5 represents the 40-50th percentile

  • 4 represents the 30-40th percentile

  • 3 represents the 20-30th percentile

  • 2 represents the 10-20th percentile

  • 1 represents the 0-10th percentile and the customers who have the lowest recency, frequency, or monetary scores.

Tip

Combine percentiles to build larger groups of customers. For example 9 and 10 together represent the “top 20%” while 8, 9, and 10 represent the “top 30%”.

To find customers who spent the most money during the previous 12 months, start with the L12M Monetary attribute in the Transaction Attributes Extended table, set its condition to is, and then specify a percentile. For example, use “10” to find customers who are in the top 10% for spend amount.

Choose the L12M monetary attribute from the Segment Editor.

The attribute will appear in your segment like this:

Find approximate RFM monetary.

Combined RFM scores

The RFM score for the customer is based on transactions that occurred within the last 12 months. The RFM score is represented as an integer between “111” and “101010”. This is a concatenated score that uses each of the individual recency, frequency, and monetary scores. The order is recency, then frequency, and then monetary.

For example, you can build an audience that contains your top 20% customers for recency, your top 30% customers for frequency, and your top 10% customers for monetary by setting the L12M RFM Score attribute to “9810” (or “9” for recency, “8” for frequency, and then “10” for monetary).

Find approximate RFM frequency for middle 30 percent.

Note

Unlike the individual RFM score attributes, the combined score for each attribute represents the lowest percentage group to include, and then includes all higher percentage groups. For example, if you set frequency to “8” it will return “8”, “9”, and “10”, or your top 30%.

This is different from the individual attributes where you could return your middle 30% by setting the L12M Frequency attribute to “5”, “6”, and “7”.

Find approximate RFM frequency for middle 30 percent.

You can access combined approximate RFM scores directly from the Segment Editor. To add this attribute to your segment, click Add criteria, select the Transaction Attributes Extended table, select the L12M RFM Score attribute, and then apply an condition.

How combined RFM scores work

A combined RFM score returns a list of customers who meet the requirement for each individual RFM score, i.e. recency AND frequency AND monetary. An RFM score of “9810” represents three individual scores. The following diagram shows individual RFM scores for individual recency (“9”), frequency (“8”), and monetary (“10”). The darker color represents the scores:

Each individual RFM score contributes to the combined RFM score.

The combined score returns customers who met each requirement. This is often a much smaller list of customers than each individual RFM score and is not a percentile. The following diagram shows the overlap and the white section represents the much smaller list of customers who met each of the individual RFM scores.

Each individual RFM score contributes to the combined RFM score.

Available conditions

The following table lists the conditions that are available to these attributes.

Note

These attributes have an Integer data type. All Integer data types share the same set of conditions. Recommended conditions for these attributes are identified with “  Recommended” and conditions with more limited use cases are identified with “  Not recommended”.

Condition

Description

is

is returns customer records with an approximate RFM score that matches the specified score.

is between

is between returns customer records with an approximate frequency, monetary, or recency score that is between the specified scores, not including the specified score.

is greater than

is greater than returns customer records with an approximate frequency, monetary, or recency score that is greater than the specified score, not including the specified score.

is greater than or equal to

is greater than or equal to returns customer records with an approximate frequency, monetary, or recency score that is greater than or equal to the specified score, including the specified score.

is in list

is in list returns customer records with an approximate frequency, monetary, or recency score that matches the scores that are specified in the list.

is less than

is less than returns customer records with an approximate frequency, monetary, or recency score that is less than the specified score, not including the specified score.

is less than or equal to

is less than or equal to returns customer records with an approximate frequency, monetary, or recency score that is less than or equal to the specified score, including the specified score.

is not

  Not recommended

is not returns customer records with an approximate frequency, monetary, or recency score that does not match the specified score.

is not between

  Not recommended

is not between returns customer records with an approximate frequency, monetary, or recency score that is not between the specified scores, not including the specified score.

is not in list

  Not recommended

is not in list returns customer records with an approximate frequency, monetary, or recency score that does not match the scores that are specified in the list.

is not NULL

is not NULL returns customer records that have a value.

is NULL

is NULL returns customer records that do not have a value.