How should operations perform user segmentation? 8 steps to clear your mind

If the thinking section gives us the angle to think, the structure section gives us the path to implementation, and the data section gives us the scientific means, then the method section gives us the efficiency of execution. It is still recommended that students who have not read the previous chapters can read them first, so that they will have a deeper understanding of the following chapters. To borrow the words of a friend , "If you have the way but not the skills, you can still seek the skills; if you have the skills but not the way, you will only be at the level of the skills."

Our "Methodology" will not specifically mention any techniques or creativity at any point in the operation process. It mainly refers to the things that must be done in the entire operation system; or the same method that will be used in many places in the entire operation process. The "user stratification" mentioned in this article is something that must be done in the operation system.

In the operation process, the role of user segmentation is very obvious. It can help us divide users into various levels and groups. Then, based on the differences between each level and group, we can formulate more accurate and targeted operation strategies.

In our operations work, we often hear the terms " user portrait ", "user stratification", and "user grouping". They seem to be somewhat similar, especially the last two, but strictly speaking, there are still differences.

• User portrait: generally reflects the objective attributes of the user. Such as gender, age, occupation, etc., are generally not transferred based on the user's subjective wishes.
• User stratification: generally reflects the user's status on the product. For example, free users, active users, paying users , high-paying users, etc. Since it is a "layer", it has a hierarchical concept and a process of progressive status, mostly in the shape of a funnel. And the user levels are generally not divided into too many levels.
• User segmentation: generally reflects the user's behavioral performance. For example, users with low frequency and high unit price or users with high frequency and low unit price may both be high-paying users, but their manifestations are different, so they belong to different groups and have different corresponding operating strategies. User grouping can also be considered as a further refinement of user stratification.

To be precise, this article should include "user stratification + user grouping", which is collectively referred to as user stratification here. In this article, we will also use an Excel table as a tool through an example to complete a user stratification process step by step from scratch.

Regarding user stratification, we need to understand the following points first:

1. User segmentation is different in different industries and may be diverse.

Those who have read our previous article "Thinking Chapter | How to Become an Operations Expert (Part 2): Operations Cube" should all remember that we mentioned the concept of users and customers at that time.

For example, in Didi Taxi , people who use the app to hail a taxi are a type of user; the driver is also a type of user; and the advertiser is also a type of user. If you want to do user stratification, you need to make a different user stratification system for these three types of users.

2. User stratification will change differently at different stages of product development.

For example, we distinguish between valuable users and general users.

In the early days, we had few products, and someone who bought twice a month and spent 200 yuan might be our valuable user.

As the company develops and our products continue to increase, you need to buy 10 times a month and spend 5,000 yuan to be considered our valuable users.

3. User segmentation requires both qualitative and quantitative analysis

As in the example above, we need to have a qualitative process for users, such as valuable users, general users, or VIPs, super VIPs, etc.; then we must quantify this, such as how much consumption is required to be considered a valuable user.

So how to use scientific methods to conduct a user analysis to determine the behavioral characteristics of each user group and complete the user stratification process? We must talk about the classic RFM user model. As shown below:

The RFM model has a long history, and its theoretical knowledge will not be elaborated here. Simply put, it uses three indicators: the most recent consumption (Recency), consumption frequency (Frequency), and consumption amount (Monetary) , and then divides each indicator into five levels according to the actual situation, forming a total of 125 types of users. Then, for the convenience of execution, the 125 types of users are summarized into 8 categories, as shown in the figure below . Finally, operation strategies are formulated based on the situations of these 8 categories of users.

One point that needs to be explained here is that the RFM model is not a product of the Internet era. In fact, it is also widely used in traditional industries, so its indicators are mainly aimed at paying users. If the user group of our Internet products is free users, we can still use this RFM model and its methods , except that the indicators are replaced by the last login, login frequency, and product usage time.

Next we will use an example to operate it again:

We now have data on 500 paying users, including four fields: (user, time interval since last consumption, consumption frequency, and consumption amount). How can we perform user segmentation and develop an effective operation strategy?

Step 1: We import or paste the data into Excel, and add three more headers: R value, F value, and M value on the basis of the original four headers. Make an Excel table like this:

(Only 10 data are selected here as examples)

Step 2: Determine the standards for the three and five levels of RFM indicators respectively.

This is a more difficult step because different industries, different products and different stages have different classification standards. For example, the consumption amount among 1,000 users is as low as 1 yuan and as high as 10,000 yuan. In most cases, 20% of users account for 80% of the amount, while 80% of users account for 20% of the amount, which is a long-tail distribution effect. Therefore, we cannot simply use the average method of maximum amount/5, or total number of users/5. The result of such division cannot represent a group with similar behavioral performance.

This mainly depends on everyone's understanding of their respective industries and actual scenario requirements. Of course, if we really have no clue, we can roughly distinguish it through a scatter plot , as shown below:

As you can see, through the scatter plot, we can see the distribution of users more intuitively (the above figure shows the distribution of users' consumption amounts). When we divide the groups, we try to group the densely populated ones together as much as possible, so that the behaviors of this user group will have greater commonalities.

It should be noted that this is not a very rigorous classification method, and requires everyone to make continuous adjustments in the actual process. If we are faced with massive amounts of data, it is best to use technical means such as clustering algorithms to help us make more scientific and accurate judgments.

In this example, we finally set five grading standards under each RFM indicator. As shown in the figure:

Step 3: Calculate the R, F, and M values ​​for each record respectively.

We can automatically calculate the R, F, and M values ​​corresponding to the record by adding an if judgment in Excel. For example, in our RFM stratified table, the R value corresponding to user 0001 is

That is, cell E3: =IF(B3>10,1,IF(B3>8,2,IF(B3>5,3, IF(B3>3,4,5))))

Let's explain this if statement:

• If B3>10, then its R value is 1, otherwise go to the next judgment;
• If B3>8, then its R value is 2, otherwise go to the next judgment;
• If B3>5, then its R value is 3, otherwise go to the next judgment;
• If B3>3, then its R value is 4, otherwise it is 5;

Using the same algorithm, we write the judgment conditions for calculating the F value and M value of each record.

F3=IF(C3>10,5,IF(C3>8,4,IF(C3>6,3, IF(C3>3,2,1))))

G3= =IF(D3>5000,5,IF(D3>3000,4,IF(D3>2000,3, IF(D3>800,2,1))))

Then, we pull down the Excel cell and finally form a picture like this:

Step 4: Calculate the average values ​​of total R, F, and M respectively.

This step is relatively simple. After we have completed all the above calculations, we will add another row at the bottom and use AVERAGE() to calculate the average of all the above rows. As shown in the figure:

Step 5: Based on the R, F, M values ​​of each record and the average of all records, determine whether the R, F, M values ​​of each record are above or below the average.

First, we add three headers, as shown in the figure:

Then, we compare the R value of each record to the average value of R. If it is < the average value, it will be displayed as "low", and if it is greater than or equal to the average value, it will be displayed as "high".

We still use the If statement for automatic judgment. Taking the above figure as an example, the "R high and low values" of user 0001 are:

• H3=IF(E3<E13,"Low","High"), similarly,
• F high and low value I3=IF(F3<F13,"low","high")
• M high and low value J3=IF(G3<G13,"low","high")

In this way, we become the following figure:

At this time, we found a problem. When we drag the cell down, E3 becomes E4, but E13 also becomes E14. Since E13 is a fixed grid number, we don’t want it to change as the cell is dragged down. We need to add the "$" symbol on both sides of E13 in the if statement.

as follows:

• R high and low value H3=IF(E3<$E$13,"low","high")
• F high and low values ​​I3=IF(F3<$F$13,"low","high")
• M high and low value J3=IF(G3<$G$13,"low","high")

At the same time, to make it more intuitive, we set a conditional format. If the text contains "high", the background color is red, and if it contains "low", the background color is green. At this time, drag the cell down again, and it will become like this, as shown in the figure:

Step 6: Classify into eight categories based on the comparison values.

 Next, we will automatically calculate our user level based on our "R high and low values", "F high and low values" and "M high and low values". Let’s first add a table header “User Level”.

This time, we are going to write a slightly longer string of IF statements, as follows:

• K3=IF(AND(H3=”High”,I3=”High”,J3=”High”),”Important Value Users”, IF(AND(H3=”High”,I3=”Low”,J3=”High”),”Important Development Users”,
• IF(AND(H3=”Low”,I3=”High”,J3=”High”),”Important Retained Users”, IF(AND(H3=”Low”,I3=”Low”,J3=”High”),”Important Retained Users”,
• IF(AND(H3=”High”,I3=”High”,J3=”Low”),”General Value Users”, IF(AND(H3=”High”,I3=”Low”,J3=”Low”),”General Development Users”,
• IF(AND(H3="Low",I3="High",J3="Low"),"Generally retain users","Generally retain users")))))))

This article describes the IF judgment statements in Excel. It is recommended that you write them yourself. It doesn’t matter if you don’t want to or can’t write them. Just save the above if statement, copy it and use it directly (just change the cell number).

Finally, as shown below:

Of course, we can also add a "filter" function to the user-level header to directly search for the users we need. You can also use different colors to distinguish different user levels, you can use your own creativity.

Well, here we have completed the whole process of user stratification by using an Excel table. The final effect of this table is that, like a program, we enter any three RFM numbers and the table will automatically jump out of the user's level. Please save this Excel sheet and use it later. It is very efficient. You can try to do it from scratch yourself. If you need it, you can leave your email address in the comment area and I will send it to you.

Step 7: Create a chart

 After completing the above six steps, we have obtained all the user records after completing the user stratification. At this time, we need to make a chart, hold a meeting, make a report, etc., as shown below:

Step 8: Develop an operational strategy

Going back to what we said above, the purpose of user segmentation is to formulate more accurate and targeted operation strategies. So, in the end, we still come back to formulating operational strategies. Our example can be seen in the figure below:

How to implement and execute it in detail is beyond the scope of this article, so I will not elaborate on it here. Please pay attention to other articles.

summary

User stratification is a very important part of the operation process, and rapid user stratification is also a necessary method for us. We divide users into layers, but the users themselves are actually unaware of it. If we divide users into levels and let them know, and not only know but also like it very much, so as to continuously guide users to move up their own levels, will the effect be good?

The author of this article @静静i黑马 is compiled and published by (Qinggua Media). Please indicate the author information and source when reprinting!

Product promotion services: APP promotion services Advertising platform Longyou Century