How to build a data operation system for a P2P platform?

It has been more than three years since I started working in the P2P industry in 2014. I am used to making some summaries of my studies, work, life, etc. in the past year at the end of the year and the beginning of the next year. The previous summaries were rather scattered. Now I want to make some summaries step by step from the framework to the details.

I have worked on a total of three platforms in the P2P industry, and my responsibilities ranged from promotion to operation , from operation to product, and from product back to operation. During the operation of the P2P platform , it was found that the mining and analysis of operational data can actually be done in great detail. Now I will use the perspective of data operation to decompose the data operation of the P2P platform from a methodological perspective, so as to build a complete data operation system.

Why do data operations?

The purpose of data operation is to accurately analyze users with certain characteristics, or even all the operating behaviors of a certain user on the platform, study and analyze their operating behaviors and make behavioral predictions, and adjust targeted operating strategies to achieve the goal of refined operations.

Data Operations Decomposition Steps

As shown in the figure below, I have broken down the steps of data operations and then explained them one by one according to these 8 steps.

Business Process

Anyone who does not understand the business process of P2P is not qualified to be a P2P practitioner.

Operational Process

The basic operating process, that is, the user's operation process, is similar for each platform. The main difference is that each platform has different functional designs for user operations or different sources of data acquisition. Here we break down the brief user operation process.

In principle, we need all the operation records of users on the platform to operate. In simple terms, it means who did what at what time and where, and what was this thing? Therefore, our first step is to obtain the user's operation records according to the decomposition steps. This is the most original and basic data requirement.

Process requirements

In addition to basic data requirements, what work requirements are generated by the corresponding departments and positions in the entire operation process? This is what we need to consider. In the general platform architecture, the operation department can be roughly divided into the following 7 departments, and the basic framework requirements involved are listed.

As for some of the data requirements mentioned in the previous process requirements, here is a brief description of the data. Please see the following table:

Regarding the process requirements, I will not list them one by one. The above is just a list of some basic framework requirements. There are too many data models that can be refined under the framework.

User Rating

Different statuses of users in the operation system can be defined in different levels, mainly according to the user status mentioned above.

Observation/Registration/Opening an Account/Recharge/ Investment /Reinvestment/Refund/Continued Investment/Withdrawal/Retrieval

After we classify users, we can extract status from the classification and make graphical statistics to obtain a user life cycle model. I extracted a data sample and made a data graph, as shown below:

As you can see from the sample data in the figure above, I have tentatively set the user's life cycle to 90 days, which is also my current definition of new users on the P2P platform (new users are those within 90 days of registration), so as to analyze the user life cycle. As shown in the chart, I divided the Y/X axis into the capital axis and the time axis. Since the user registered, such data has been available. After importing the data into the software, the result is as shown in the figure.

From the data in the example graph, I draw the following conclusions:

• The period starts from the user's registration time and ends with the first recharge time. The user took about 10 days to make the decision.
• From the time of the first recharge to the end of the first investment, the user has about 5 days to make the decision.
• Since the first payment, the user has carried out withdrawal operations and continued investment operations in chronological order. It indicates that the user has experienced the withdrawal time limit of the platform and then continued to invest.
• The shorter the interval between the user's recharge and investment decision-making time (timeline), the higher the corresponding funds, which means that the user is gradually entering the sedimentation period.

The decision-making behavior of corresponding users is influenced by many external factors, such as: brand events, marketing activities , customer service follow-up development, etc.

Strategy Adjustment

Extract user classification data and adjust operation strategies after analysis.

Here are two common data applications for illustration:

1. In the general traffic promotion channel , how to judge and analyze which new users are relatively more valuable to develop under certain equal conditions.

• The higher the difficulty of user decision-making, the greater the development value. For example, the registration process is relatively simple, but opening a bank deposit account with real name is much more difficult, and users need to fill in highly private information such as their ID number and bank card information, which makes decision-making very difficult for users. Next comes the first charge and first investment.
• The decision time from registration to account opening, and the decision time from registration or account opening to first deposit and first investment. Before the advent of highly accurate anti-fraud (user’s fake account) functions, these two types of users with shorter decision-making times were most likely guided by marketing activities and channel -targeted CPS activities in making decisions. Relatively speaking, the longer the time for these two types of decisions is, the more it shows that the users are genuine to a certain extent, have a certain observation period for the platform, and make decisions only after careful consideration.
• In the CPS promotion channel, for users who are attracted by certain reward returns, the initial investment amount of most converted users comes from the activity threshold of the CPS promotion channel. If the initial investment amount of users from this channel is greater than the channel activity threshold, it means that these users are relatively more valuable to develop. In addition, the higher the initial investment amount and term, the greater the development value.
• Before entering the sedimentation period, users have a certain trial experience period. During the decision-making time span of this period, users who reinvest without receiving payment have more development value than users who continue to invest after receiving payment. In addition, the higher the amount and term of reinvestment and renewal investment, the greater the development value.
• Under the premise that the user has zero pending payments and the balance of the user's escrow account may also be zero according to the formula, users who still log in and visit the account are relatively more valuable for development. The higher the access frequency and browsing time, the more valuable the user is to develop.

Then, among the data information obtained from various aspects, gradually analyze the users who are relatively more valuable for development. This will help adjust the work of the customer service department and promotion department to target specific groups and strategies.

2. During the non-activity period and the activity period, excluding the influence of other external factors, users with different investment capabilities will have to reach what expected value of the returned funds before they can make a decision to withdraw or reinvest.

As shown in the data sample, if the user receives payment for 15 consecutive days and the average daily payment amount reaches ≥759.9 yuan, the payment renewal operation will be carried out. The minimum payment renewal operation will be performed when the cumulative payment reaches 518.17 yuan. Therefore, we can preliminarily infer the user's decision to reinvest based on the data. The expected value of the decision requires the repayment amount to be ≥ 518.17 yuan before the operation will be executed.

Among the user's repayments, there was a repayment of 973.76 yuan, and a recharge of 50 yuan was carried out on the same day before the investment operation was carried out. This is a very strange behavior, just like the rounding-up mentality that many of us have. The psychology of rounding up is common in the marketing methods of some e-commerce companies . If there are friends who have studied behavioral psychology, please provide some information about the psychology of "rounding up".

Next, we make the following hypothetical questions based on the data for your reference.

First look at the following picture:

• It is assumed here that the user investment ability levels are set at 5 levels (as described above), and the user in the previous example is at a low level of investment ability. By analyzing all users according to the data model, we may find that when users of different levels make renewal decisions, their expected decision values ​​can be roughly estimated. Can this estimated data be used more accurately by the customer service department to track and return visits to customers?
• Similarly, when users make withdrawal decisions, for users at different levels, when the returned funds reach what expected value will the user execute the withdrawal operation?
• Can the total investment amount of each user on the platform (commonly known as warehouse) be analyzed through data models to measure the comprehensive capabilities of the platform? What is the range of users' psychological expectations for building positions on the platform?

The above issues should be left to platform operators who have large amounts of data and certain technical capabilities to solve data acquisition problems.

Guidance

Each operation link (i.e. the link of user classification guidance) is guided from the aspects of content, interaction and vision, and the guiding effect is observed.

The data in this aspect include page visit data (number of visits, repeat visits, visit depth), information arrival data, conversion data (activation, awakening, conversion), etc. Basically applicable to: customer service department, promotion department, and operation department. Applicable to: event planning positions, copywriting positions, product positions, and design positions.

Multi-dimensional trade-offs

From each adjustment of the operational strategy, we dig into multi-dimensional in-depth data and weigh the data accuracy.

Perhaps the data demand here is more in event planning and customer service positions, such as event participation, user activity, information reach rate, activation conversion rate , increase in decision-making behavior, etc.

Variable influence

For each operational data variable, focus on analyzing the multi-dimensional impact values ​​involved in the data variables.

This part mainly includes: user withdrawal behavior, abnormal user operation behavior, abnormal user activity, abnormal user decision-making habits, etc. It is mainly used for risk warning and behavior prediction, and the main data is suitable for customer service department, brand department, and operation department. Applicable to: customer service positions, negative information monitoring positions, event planning positions or transaction management positions.

Okay, let’s stop here for this sharing. It’s true that my mind is not active in such cold weather and it’s hard to write. There are some contents in this sharing that are not explained in detail. I will share them step by step later. Or I will share my current data modeling ideas after I have further demonstrated them. I am currently organizing a set of data models, which may be around 100 in number.

The author of this article @Mr路人丁 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