Why can’t user portraits be implemented?

User portraits , as an effective tool for outlining target users and linking user demands with design directions, have been widely used in various fields; especially in today's big data era, user portraits can also help companies to accurately position themselves to a great extent.

When talking about user portraits, I believe everyone is familiar with it.

This is the persona that PM likes to use in daily life, because it allows them to be more specific and convincing when discussing requirements.

So how do user portraits come about? Besides being somewhat useful when discussing requirements with product managers, what else can it do? Please listen to me one by one.

1. What is User Portrait?

The premise of user portrait is labeling. User portrait is composed of various labels, so the bottom layer of user portrait is labeling. As long as the labels are well-designed, the user portrait will never fail.

The practice of most companies is for the data team to find distinctive labels through data statistics, which generally include labels such as user attributes, user value, transaction attributes, interaction attributes, product preferences, and behavioral preferences.

Then form a user portrait based on the tags. However, this type of portrait is difficult to implement with specific operational strategies. For example, if someone is aged 25-30, male, and lives in a first-tier city, how should the operational strategy be implemented?

Because these labels are created by data colleagues rather than operations colleagues, they are often divorced from business realities and difficult to implement.

Over the years, even for a company of Alibaba’s scale, only the product preference labels have been applied relatively well. There are very few labels that are truly useful and implementable for a product, and a dozen or so are actually enough.

If you really want to implement it, you should not focus on the tags and think about how to apply them. Instead, you should think about the business process first, and then think about how to find tags to identify users.

Use behavioral labels as the main logic (user stratification) to refine the first step, and then use attribute labels to further subdivide different groups (user grouping) based on the main logic.

For example, for a transaction product, user stratification can be divided into users waiting to be activated, activated users who have not traded, activated users who have traded, and old users who have traded.

Then, we can make detailed user groupings under each layer, such as dividing activated but untraded users into several stages: somewhat interested, looking for targets, and wanting to buy.

Then, some behavioral tags and other tags are combined to identify specific groups of people. Users at different stages have different needs. We will introduce the main products that they are interested in, the main popular products that they are looking for, and timely remind them to pay attention to product updates if they want to buy.

In addition to the main logic, attribute tags can be used to optimize the copy. For example, if the target audience is young people and interested, then the copywriting can be more interesting. If you are older and have an interest, the copywriting should be more formal.

The above is a feasible version of the user portrait, which will be used for user stratification and user grouping. Next, we will sort out the relationship between them and tags.

2. What are user stratification and user grouping?

The best scenario for user tags is the recommendation system, which can deliver precise push notifications through different tags, meet the needs of different people, and improve business conversion rates.

At the level of refined operations, user tags serve to stratify and group users. Without user portraits/user tags, it is impossible to effectively stratify and group users.

Why do we need to do user segmentation?

User segmentation is a method used when the number of users is very large and difficult to manage. It is more effective to segment products with strong business attributes because they aim to increase conversion and achieve growth. User segmentation uses different operational strategies to improve conversion and achieve ten different faces for each user.

The common logic of user stratification is to stratify from low to high based on the degree of relevance between users and the core business of the product, and from low business volume to high business volume. After the layers are divided, users are grouped.

Let’s first look at a simplified MVP version of user stratification + user grouping. If you are a micro-businessman, how would you stratify and group your WeChat friends?

Make stratification from difficult to convert to easy to convert, and make different groups under each stratification. Using this as a model, we can create a prototype for user stratification and user grouping for all trading products.

I believe that you, as a smart person, have thoroughly understood what user stratification and user grouping are.

3. How to implement user segmentation?

User stratification is a pyramid model. Each layer is completely independent. The layers have specific business distinctions, and the things users do are different.

User stratification is based on artificial division of some stages and standards for each stage.

The purpose of stratification is to hope that users convert to higher levels. If they cannot convert upward, then your user stratification is done wrong.

The value of product stratification with a short decision-making process is not very meaningful, while product stratification with a long decision-making cycle is more valuable.

For products where user decisions only take one day, user segmentation only needs to be done into: new users + old users. The operation for new users is to hope that they can trade. The operation for old users is to hope that they can continue trading.

User groups generally overlap, and a user may belong to both group A and group B. Grouping is done based on stratification, such as gender grouping and age grouping in e-commerce.

The RFM model is commonly used for clustering and is widely used in the market, but in fact this model may not be very effective. The RFM theory is as follows:

Let’s take a look at the RFM model actually used by Meituan:

The core logic of Meituan RFM is as follows:

Why is it said that the RFM model may not be effective? It originated from the retail industry and may not be applicable to the current C-end products on the Internet.

• For users whose activity is decreasing, push something they are interested in.
• For users who have been browsing but are reluctant to place an order, give them some coupons
• High-quality users have lost, give them some return gifts
• If you are worried about your core high-quality users, you can give them value and respect by offering points or membership.

…

Are these good operating practices? I'm sorry, these are just academic talk and not practical enough. You can find an intern to do the job.

The key to doing user segmentation and user grouping is not to build a good model in advance, but to look for answers with questions.

Excellent operational actions should be user reach that meets user needs, provided that the user’s contradictions in the product are discovered, as shown in the figure below.

There may be no magic bullet in this world. No strategy or action can be achieved overnight. It is normal to do poorly at the beginning, but you must test and iterate frequently.

The most terrifying thing is not that we didn’t do it well, but that our users have changed, the data has changed, the product life cycle has changed, but the operational actions remain the same old way.

The topic has gone too far, so user portraits, user stratification, user groups, user tags, do you understand what they are, smart as you are?

Author: Xiaohe Operation Notes

Source: Xiaohe Operation Notes