How to do effective “user segmentation”?

The establishment of an effective user segmentation probably involves three steps.

My uncle opened a ramen restaurant in the city and asked me to help. I happily agreed.

After working for a few months, I became familiar with the customers coming and going in the small restaurant.

Boss Zhang is a regular customer here. He comes three or four times a week, and sometimes brings a group of friends with him, and orders lavishly.

We are especially grateful to this guest. When there are many people, we will give him two extra side dishes, round off the amount when he settles the bill, and prepare a box of local specialties for him to take home during festivals. Over time, we became friends and couldn't live without each other.

Brother Li's son especially likes our lamb skewers, so Brother Li often brings his son here, but recently he always comes alone. He said that his son just went to middle school and couldn't come because he got out of school too late.

When checking out, I gave a few skewers of lamb to Brother Li and asked him to pack them up to take back for his children. Brother Li was very happy. From then on, every time he came to eat, he would order 10 more skewers of lamb to take away.

Miss Yang is a white-collar worker in a nearby company. She is young and beautiful, from the north, and likes to eat noodles. Because I am busy with work, I am always in a hurry every time I come here, just eat something and leave. I felt sorry for her, so I recommended our takeaway service to her. She ordered a few times and thought it was good, so she recommended it to her colleagues, which brought us a lot of business.

…

With my familiarity with every customer, I can always provide different customers with the services they want . Before I knew it, the business became booming. My uncle opened several branches, and the business spread to the provincial capital.

I rose from a waiter to a manager, but problems arose - before, there were only a few dozen customers and it was easy to treat them differently. Now, the business is bigger, with at least tens of thousands of customers. It is impossible to get to know them all, let alone provide one-on-one service.

When I was feeling very worried and at a loss, my uncle's eldest son graduated from college and came home and gave me a suggestion.

If you can’t remember it with your head, why not use a computer?

Great idea! I immediately called together dozens of store managers under me and asked them to organize all the customer information since the opening, including how often they came to dine, how many people ate together, what dishes they ordered, how much they spent, etc. I also told them to pay attention to the collection of customer information in the future, and it would be best if it could be standardized.

A few days later, the store managers started sending me some customers' consumption data. Following the method taught by my eldest cousin, I classified the data and organized them into different groups. Among these groups, I found thousands of Boss Zhang, Brother Li and white-collar Miss Yang.

I immediately designed a set of standardized service and sales methods for different groups of people, printed them into booklets, and distributed them to all store managers, asking them to follow them.

Sure enough, before long, store managers reported that their performance had increased and customer satisfaction had improved a lot.

My older cousin who went to college told me that this method is called user stratification.

The “Tao” of User Segmentation

What exactly is user segmentation?

Many user segmentation methods can be found on the Internet. The simplest one is the 28 segmentation, the common ones are the user value pyramid , user life cycle segmentation, AARRR model , and slightly more complex ones include the RFM model .

These things can be considered as the "techniques" of user segmentation. The method structure is clear, but it may not be able to adapt to the complex and diverse specific needs.

Take the RFM model as an example. This simple response model, which grew up in traditional industries, was originally used for direct marketing and to calculate the break-even point. In the Internet environment with very different conditions, RFM can be described as a classic in terms of thinking and a revolutionary in terms of method . With all due respect, many articles online discuss the RFM method in a very eloquent way and call it "user segmentation." This is more like making up questions with answers in mind, rather than looking for answers to problems.

If the technique deviates from the Tao, it will surely decline.

So what is the “Tao” of user stratification?

Simply put, it means treating different users differently rather than ignoring the differences between them .

To go further, it means providing targeted marketing and services based on users' different consumption tendencies and preferences to polarize user value .

Just like the case of my uncle's ramen restaurant, after fully understanding and analyzing each customer's consumption tendencies, I can always provide targeted services and consumption suggestions to them, making it convenient for them while tapping into their individual consumption potential.

If I take it to the extreme, I can even remember who doesn't want chopped green onions and who has a heavy taste for salt, thereby further improving customer satisfaction.

The same is true for any commonly used user stratification model, such as the value pyramid model:

Different users use the product to different depths, so their feelings and needs are different.

The purpose of user segmentation is to explore the potential needs of individuals and use targeted incentives to realize these potential needs, thereby accelerating the upward flow of the pyramid.

If you think the above content is too basic, then I’m glad you can come to the main text below.

User stratification: how to control it with Taoism

Let me first talk about a small principle.

On the second day after I joined the e-commerce company, my boss asked me to prepare for the upcoming big sale and do user segmentation for SMS push notifications.

Even if you have expertise, you can't be too arrogant when you take on your first task after joining the company. I enthusiastically asked BI to pull data, worked furiously on the spreadsheets, diligently divided all users into more than 20 layers, and confidently emailed my boss.

The next day, the boss ignored me and directly gave the task to someone else. My boss told me that out of the more than 20 layers I made, only 5 were necessary to promote, and the other dozen or so were not unreasonable, but they were just nonsense .

What does it mean?

For example, in my stratification, there is a "customer unit price is less than 100 yuan, the number of historical purchases is less than or equal to 3, and there is no visit in the past 90 days". For this stratification, my suggestion is "low-quality churn users, it is recommended to abandon them."

Boss: What should the users who are about to give up show me? Am I free?

For example, there is another stratification in this layer: "The number of historical purchases is greater than 10 times, only purchased beauty products, and visited in the past 7 days." My suggestion is "High-quality users, single purchase category, can push coupons for strong categories on the platform (luggage and clothing) for cross-selling"

Boss: I want to see the most direct output of the big sale. Is this related to the big sale?

So what is meaningful layering?

For example, "customer order value is greater than 500 yuan and there has been no visit in the past 90 days" is for recall; "customer order value is greater than 500 yuan and there has been no visit in the past 7 days" is for activation; "added to shopping cart but not purchased in only 7 days" is for conversion .

Implementing these stratifications through push notifications and following up with big promotions are the most urgent tasks at this moment.

Therefore, the correct way to think about user segmentation should be to find different opportunities for different users based on business needs . The "differential treatment of users" mentioned in the previous article is not wrong, but it is only user stratification in an academic context .

When doing user segmentation, there is no need to classify all users into the segmentation. Just give a general outline, just like catching fish in the sea or digging for gold on the ground, whatever you catch is valuable.

It’s the same old saying, we can’t look for problems with answers and methods, but we should look for answers with problems .

I've been talking too much, let's get to the point. How to do user stratification?

The establishment of an effective user segmentation probably requires three steps:

1. Find the conflict

The so-called "conflict discovery" means discovering behavioral conflicts when users use products.

What is conflict?

Interested but not understanding, understood but not buying, bought but not repurchasing, repurchasing but buying less, buying a lot but suddenly stopped buying.

Conflict means that users encounter some kind of obstacle during use , which may be information asymmetry or the product may not be good enough, resulting in the user's actual needs not being realized.

Only when there is conflict can it be resolved. If there is a lack of information, we can communicate. If the product is not good, we can improve it. Once the conflicts are resolved and obstacles are removed, and the users' demands can be realized, the users in this stratification can be operated well.

Therefore, conflict is the premise of user stratification. Layering without conflict has no underlying demand and is meaningless.

Let’s take a simple example:

Users who purchase frequently suddenly go missing, users who have been buying crazes recently have become less active, users who have been browsing are hesitant to place orders, receive coupons but don’t buy, add items to cart but don’t buy... These are typical behavioral conflicts, and they all require the establishment of layers to resolve conflicts through operational means.

To put it more simply, when some of the user's data indicators are very high, but the data indicators that should be positively correlated with them are very low, it means that there is a conflict.

On the contrary, if some users have a high average order value, buy frequently, and are very active, it means that their consumption is smooth and there are no obstacles. In other words, some users’ indicators are all on the edge, which means that they may not have a clear demand for the product. In these cases, stratification does not hold true and no operations are required.

Some people may say that high-quality and stable users also need to be managed, and we must do a good job in preventing loss! This is the second thing I want to talk about.

When you have worked in an industry for many years and have a keen business sense, you can not only detect conflicts, but also predict them.

You will know at what time and under what circumstances what kind of users are at risk of sinking or losing. The greater the risk, the more operations are needed. If the risk is smaller, you need to consider the input-output ratio and then decide whether to divide the operations into layers.

Take high-quality and stable fan users as an example. If the fan user group of a product has been extremely stable for several years with almost no churn, then it would be unnecessary to have an anti-churn mechanism.

On the contrary, if the users at the top of the product pyramid are like "soldiers who come and go but remain unchanged", with different faces coming in and out every year, then it is very necessary to create a VIP mechanism and find ways to establish a user exit wall to slow down the loss.

In addition to the loss of fans, potential conflicts are also hidden in many places. If you want to dig them out, there is no universal operational methodology and you can only rely on accumulated business experience and sense. The logic is very simple: when experience tells you that certain groups of people are likely to have behavioral conflicts in certain situations, pull out this group of people and stratify them, and find ways to avoid the occurrence of conflicts.

In summary, the existence of conflict or potential conflict is the first condition for establishing a hierarchy.

2. Find leverage points

Value arises from the resolution of conflict.

User segmentation is to resolve the different conflicts among different segments. To resolve these conflicts, some operational means are needed, such as Push, coupons, membership mechanism, etc. When these means can be supported by existing resources and can leverage considerable benefits with limited investment, it becomes a leverage point .

For users with declining activity, push them things that interest them; for users who have been browsing but are reluctant to place an order, push them a coupon; for high-quality users who have left , send them some return benefits to welcome them back; if you are worried that users at the top of the pyramid may leave at any time, give them enough value and respect, and make them feel that if they leave, the status they have accumulated before will be difficult to find on other platforms.

Are these leverage points? Sorry, this is all just paper talk.

The leverage point I want to emphasize is not strategy and means, but goals and resources .

Take the cross-selling stratification that my boss dissed me about earlier as an example:

1. The goal of a big promotion is to boost sales. For loyal users of the beauty and cosmetics category, repurchase of beauty and cosmetics is the most direct way to generate GMV. Cross-selling of categories is difficult and risky, and requires long-term observation and resource commitment.
2. Big sale resources are large quantities of discounted products and coupons that burst out instantly, which are suitable for conversions. Cross-selling requires long-term commitment to products and discounts in a certain category and is not something that can be achieved overnight.

In summary, there is indeed a conflict in this stratification (high-frequency, high-activity, high-quality users VS those who have never purchased the platform's main categories), but the resolution of this conflict does not fit the current business goals (big sales to boost GMV) and there is not enough resource support, so this stratification cannot be established.

So, again, user segmentation is just a means, a segmentation opportunity point to solve problems towards current business goals. User stratification that is separated from business goals and resources can only serve as an academic methodology.

3. Generate benefits

Efficiency is not the income, but the input-output ratio.

Generally speaking, the more intense the conflict or potential conflict within a layer, the greater the benefits of resolving the conflict. For example, the conflict between “high-quality users vs. churn” is more serious than that between “ordinary users vs. churn”, and the corresponding recovery of high-quality users can also bring more outstanding benefits to the platform.

The cost investment varies from person to person. The same method may mean different costs for different platforms.

For example, coupons mean a 1:1 subsidy for B2C and C2C e-commerce companies. For self-operated e-commerce companies, since they have a certain degree of pricing power, there are many ways to use coupons, and the cost will not be as high as it seems. For example, JD.com currently lets you draw a coupon every time you place an order. The amount of the coupon is very high, but the products they are for are all self-operated products with inflated prices.

The comprehensive benefits and costs are actually the comprehensive intensity of conflicts and the investment of leverage points . You can then make a decision as to whether this stratification is valid and whether it is worth operating.

Mind Map

Now it’s time to do some practice.

First, think carefully about the business goals set by your boss, use BI to pull out key user indicator data for observation, and combine business experience to see which users have conflicts in their behaviors. Determine whether resolving these conflicts can help achieve current business goals and whether there are enough resources to resolve these conflicts. If the answers are all yes, you can circle out the user groups with sufficient operational cost-effectiveness and use them to do things.

If you want to do strategic stratification, that is, things like membership mechanisms, you should focus more on potential conflicts. Think back to your interactions with users over the past few years. Which users always become less enthusiastic or leave you at what time? Seize these people, this time, and these potential risks, seek leverage points, and form product mechanisms.

Source: Operation Dog Oliver