Let's take an example of Internet catering ~ to prove how to analyze the RFM model:

How to analyze the basic attributes of the user through the takeaway order data;

The user's orders have ordering addresses, through the statistics for the ordering address, we can query the distribution of users under different combinations of conditions, and even know where the users who like a certain dish are. Similar user data mining can also be based on the composition of reordering, reordering users cross-platform use, gender composition to do a more refined analysis. It is worth noting that the differences between data platforms are still quite large, which is conducive to making different marketing strategies for different platforms.

The above basic user attributes are still not enough for refined operations. Because this information can not help you solve the following four questions -

1. Who are my important value customers, what are their characteristics?

2. Who are my customers that I need to focus on keeping in touch with and what are their characteristics?

3. Who are my key development customers and what are their characteristics?

4. Who are my key retention clients and what are their characteristics?

2. How to divide users into groups through RFM model to realize refined operation

RFM model is a widely used customer relationship analysis model, which mainly distinguishes customers by user behavior.

F = Frequency Consumption Frequency

M = Monetary Consumption Amount

Needing to understand the definition of the above three metrics in detail, Baidu will sub-divide the dimensions into 5 more copies, which will be able to sub-divide into 5x5x5=125 categories of users, and then according to each category of users precision marketing ...... Obviously 125 categories of users have exceeded the calculation of the ordinary human brain category, not to mention customized marketing strategy for 125 categories of users. In practice, we only need to do a two-way split for each category, so that we still get 8 groups of users in 3 dimensions.

Important value customers (111): recent consumption time is close, consumption frequency and consumption amount are very high, must be VIP ah!

Important Keeping Customer (011): recent consumption time is far away, but consumption frequency and amount are very high, indicating that this is a period of time did not come to the loyal customers, we need to take the initiative to keep in touch with him.

Important development customer (101): recent consumption time is more recent, consumption amount is high, but the frequency is not high, loyalty is not high, very potential users, must focus on the development.

Important retention of customers (001): recent consumption time is far away, consumption frequency is not high, but the consumption of high amount of users, may be about to lose or already to lose the user, should be based on retention measures.

3. How to build RFM model on BDP Personal Edition to help users divide into groups

At this point there may be a friend who asked, gosh, you have this three-dimensional model, and there's no way for me to build a form with BDP. So what we need to do is to 2D the 3D model, we will cut a piece out of the R domain (i.e., do the analysis in the last 30 days of users with repurchase), squash it and you will see.

The representation above is perhaps still too academic to be simple

Step 1: start by picking out the users who have repurchased in the last 1 month.

Second step: the average out-of-pocket amount of repurchase users in the last 1 month to do the vertical axis.

Step 3: The purchase times of the repurchase users in the last 1 month do the horizontal axis and generate a table.

Step 4: You will need to redline this table yourself.

Horizontal red line, represents what you think the average guest who comes to dinner should spend how much money per meal, I set the value of 25 yuan here, call takeout 25 have not paid to me is a low consumption amount (low M) users.

The vertical red line, representing how many times you think repurchase guests, is your high-frequency users. Takeout ordering has a high turnover rate, and a user who can order more than three times a month at a store is a high-frequency user to me.

In this way, the RFM model on the personal version of BDP has been established. What is the use of this RFM model in practice? For example

For example, if the conversion rate of group text messaging to circle users is less than 1%, you can use RFM to do an analysis, and only select users with high R-value (users who have spent money in the last 2 weeks to the last month), the conversion rate can be increased from 1% to 10%.

This also means that the previous $6/order will drop to $0.6/order. The bosses are willing to spend 600 yuan to send text messages to 10,000 users and get 100 orders, or are willing to spend 48 yuan to send text messages to 800 people to get 80 orders, I believe we will choose the latter.

And the overall RFM differentiation, can help the shopkeepers for different users to send different text messages, the beginning of the text message is to use "long time no see", or with "congratulations on becoming a VIP", it depends on the time important to keep the customer or an important value The user. Only to be able to distinguish between users, in order to move towards the refinement of the operation.