Those who have read my first article, Getting Started with Internet Finance Operations (Part 1): An Article to Understand the Six Types of Internet Finance Products, should know that the first article mainly introduces various types of Internet finance businesses, players, profit models, core capability requirements, etc., which is actually a process to help everyone Know Your Product .

Understanding the product you want to operate is the first step in operation, otherwise you will never be able to operate it well.

In this article, we will talk about another thing: Know Your Customer. Know who the target users of your product are, where to find them, what they care about, and what can impress them. Only in this way can you become a calm and experienced operator.

Is user portrait important?

Without further ado, let's discuss a few questions together:

1. Which type of your users contributes the highest AR PU? What are their characteristics? Where can we best reach this type of users?
2. Which type of users retain the best? What made them stay? What else do they need?
3. Which users will be first open for testing of this new feature?
4. The activity of APP has been decreasing during this period. Why?

If you can answer the above four questions without guessing, I believe you basically have a 1:10000 scale user portrait map in your mind.

So looking back, the above daily operational issues that seem to have nothing to do with user portraits all require a deep understanding of users.

1. Five-dimensional user portrait

Each industry has its own unique user portrait. E-commerce may focus on consumption capacity and brand preference, O2O may focus on regional attributes, and the financial industry naturally focuses on the user's asset status. Since the Internet finance industry combines the characteristics of the financial industry and the Internet industry, the user portrait will be more multidimensional. The picture below is the user portrait of the Internet finance industry that I summarized. I call it the five-dimensional user portrait .

This five-dimensional portrait describes the user from five dimensions: basic information, financial attributes, consumption attributes, interest attributes, and social attributes. Through these descriptions we can get a general explanation of the user.

But who is the user at this time? Where is it? How to find him? We still can’t answer this question because we need a second step: cluster analysis of users on this basis to create a typical user profile for the product.

(Example picture)

The purpose of a typical user portrait is to outline a representative user, so that when we read the portrait, the person's voice, smile, joy, anger, sorrow and happiness will immediately appear in our minds, and it will feel like a friend around us.

In addition, it is important to ensure that the portrait has flesh and blood while ensuring that it has a skeleton support. I call this skeleton the user DNA indicator.

Any indicator that can distinguish a certain type of user from other users can be defined as a DNA indicator, such as average order value, frequency of financial management , risk preference, etc.

Based on the DNA indicator, we can build a user classification model, which can help us solve most of the problems in operational work, but remember to control the number of DNA indicators in a model to about 5.

(Classification model diagram)

With this model, subsequent operations have been 50% successful. Next, I will give an actual case to help everyone understand how to use user portraits in actual operations.

2. User portrait operation case

This is a trading case I encountered when I was operating MGM.

MGM is a common customer acquisition model that recommends friends to register, and the key factor of the model is the recommender.

In the early stage of operation, I analyzed all recommenders according to the five-dimensional portrait, but soon encountered a problem: the portrait is a mean description of all recommenders, and according to this description, the characteristics of all users are extremely similar.

In addition, the contribution value of different recommenders is different. What characteristics of users do we want to find the group with the highest ARPU?

In order to solve these two problems, I extracted the three most core indicators from the five-dimensional portrait of the recommender and built a recommender model , dividing the recommenders into eight categories. For example, the red recommender in the figure below, I call him the "forward recommender", which is similar to a forward on a football field.

There are 5,000 such recommenders, accounting for 20% of all recommenders, but the loan amount from users they recommend accounts for 79%, which is very consistent with the "80/20 principle". Whether they can be operated well directly determines the success or failure of the operation.

In June 2017, our business funding needs suddenly increased, and we needed to add an additional 100 million yuan in funds within three days. This was a typical difficult task with tight time and heavy workload.

Conventional strategies failed to accomplish the task and planned activities were not launched in time. After analysis, we decided to target this wave of forward recommenders. That afternoon, a limited-time commission rebate text message was quickly sent to these 5,000 people. In just three days, a total of 1,000 people participated in the event, bringing in 120 million yuan in funds, completing the task quickly, accurately and ruthlessly.

The key to the success of this activity was actually to have a deep understanding of the users before the tasks appeared. Without the five-dimensional portrait and eight-category recommender model, it would be impossible for us to accurately locate users and estimate effects in a short period of time. In fact, this applies to all operational work, and this is also the inherent strength of operations.

Finally, let’s summarize the three stages of using user portraits:

1. Describe the target user based on the five-dimensional model and form a user profile;
2. Extract 3-5 key indicators based on operational goals and cluster and segment users;
3. Carry out targeted operations and review for each type of user, and finally summarize the right strategy.

The currently popular personalized display, customized operations, and smart recommendations all follow the same logic, except that the data source has been augmented with behavioral trajectory data, making operational actions more real-time and dynamic.

Source: Beijing Soldier