Determine the RFM division dimensions and hierarchical standards:
These data need to be divided into levels according to the three dimensions of Recency, Frequency and Monetary. Generally speaking, it is more common to divide these three dimensions into four levels.
This will generate 64 different customer groups (4x4x4). Of course, we can also use three levels, which will generate 27 different customer groups (3x3x3). However, it is not recommended to use more than 4 levels, otherwise the difficulty of using the segmentation model will increase a lot.
For ease of understanding, the customer transaction data we use here is simplified, with only three customers and relatively few transactions. In fact, the data that customer service personnel have to deal with is much more complicated than in the example, so it is generally necessary to use Excel spreadsheet formulas or other professional software to analyze the specific scores of Recency, Frequency and Monetary.
(1) Recency score
Assuming that the date we collected this data is September 1, 2019, to calculate the number of days since the last transaction, we can extract the number of days since the last transaction date of the customer. The data of other customers can also be extracted in the same way, or by using relevant Excel formulas. Therefore, we calculated that the number of days since the last transaction of Ian, Janet, and Raj were 21 days, 3 days, and 5 days respectively.
Based on this data, we set the classification criteria for each level of Recency as follows.
R1 level (recent): <15 days
R2 level (closer): 16~30 days
R3 level (longer): 31~45 days
R4 level (longest): ≥46 days
Therefore, Ian’s Recency score level is R2, Janet’s Recency score level is R1, and Raj’s Recency score level is R1.
(2) Frequency score
Frequency refers to the number of transactions. For cross-border e-commerce, we need to decide whether to consider all visits to the website or mobile application as frequency or only the frequency of successful transactions/conversions. In this case, we only count the number of transactions for each customer and use it as the frequency. Therefore, Ian has 1 transaction, Janet has 5 transactions, and Raj has 2 transactions.
Based on the actual store operation and the nature of the goods, the Frequency score range is set as follows.
F1 (most frequent): >10
F2 (more frequent): 5~9
F3 (less frequent): 2~4
F4 (only once): 1
Therefore, Ian’s Frequency score level is F4, Janet’s Frequency score level is F2, and Ra’s Frequency score level is F3.
(3) Monetary score
The spending amount is the total spending amount of each customer in the selected time range. It is calculated by summing up the transaction amount. In our case study, Janet spent the highest amount of $103, followed by Raj who spent $65, and Ian who spent the lowest amount of $12.
Based on the actual store operation and the nature of the goods, the Monetary score range is set as follows.
M1 level (highest): ≥180
M2 level (higher): 100180 (inclusive = 100)
M3 level (lower): 30~100 (inclusive = 30)
M4 level (lowest): <30
Therefore, Ian’s Monetary score level is M4, Janet’s Monetary score level is M2, and Raj’s Monetary score level is M3.
Customer Ian belongs to level R2 in terms of the number of days of the last consumption, F4 in terms of consumption frequency, and M4 in terms of consumption amount, so he belongs to RFM level 2-4-4. Customer Janet belongs to RFM level 1-2-2, and Ra belongs to RFM level 1-3-3.