Practical analysis of user operations and behavioral data insights!

What exactly is user operation ? What abilities and qualities are required to become a user operator?

What is user operation?

As the name suggests, user operations refer to any operation and strategy around the user's total LTV:

Some people think that organizing activities and building communities are what user operations should do;

Some people think that user operations should focus on user systems, conversions, and data monitoring;

Some people also include channel delivery operations into user operations. In fact, I think these can all be called user operations, because these tasks will affect the user's entire LTV cycle. So what abilities and qualities are required to become a user operator?

Where do users come from?

1. Channel distribution function division

There are two ways for users to enter the web and app: paid and free. Usually, the user diversion entrance is opened on the web, app and some traffic distribution platforms. According to the functional division, it can be roughly divided into five channel operation functions: SEO, SEM, ASO, ASM, and DSP (information flow).

1. Payment channel settlement method

Usually, the settlement strategies of SEM and DSP during delivery can be roughly divided into: CPC, CPA, CPM, CPD and other settlement methods. Among them, CPS and CPA are beneficial to buyers, while CPC and CPM are beneficial to sellers.

1. Resource docking

Like stock trading platforms, different buyers and sellers will also connect resources through intermediaries. Ad Exchange connects the buyer (advertiser - the one who places advertisements, such as our App) and the seller (the owner of the traffic advertising space of the SSP, such as iQiyi) of advertising transactions.

Through the above methods, users have come to our products.

(Internet distribution channels)

Product Construction

User stratification & user label system construction

User tags are a prerequisite that is very important in user operation strategies. If certain user behaviors can be automatically tagged, then a good foundation can be laid for subsequent vertical refined operations.

The first type of label can be used to stratify users according to the business type that the company values ​​most, such as the profit of an e-commerce company, the user recharge amount of a live broadcast company, and the handling fees of a transaction company. If we further split it, it could be dimensions such as transaction amount, profit, and recharge amount.

The second type of tags can use user attributes to stratify users, such as user age, occupation, gender, etc.

The third type of tags can be based on user behavior on the Web or App, such as browsing, viewing, registering, ordering, paying, and repeat purchasing. However, this project has high requirements for product placement, data, and background support.

There are more user tagging systems, such as tagging users based on their LTV life cycle and tagging users using the RFM model, which are popular traditional Internet operations.

So how do we use user tags after we have them? This will be explained in detail in the following operational strategies.

(User stratification & user label system construction)

Product feedback mechanism

Establish a "not interested" function. Currently, many product managers lack good communication with users when making pages, resulting in a lack of understanding of user preferences. Therefore, it is crucial to set up a user suggestion and complaint box and a "disinterest" function on the product side.

User incentive system & user growth system

This is a topic that has been talked about for a long time. For a long time, if an App did not have this thing, it would not be called an App. In fact, this user growth system is divided into different types of products and should be made in different ways.

For example, the revenue-generating behavior of high-frequency products such as Moji Weather is mainly concentrated on advertising display, so it is crucial for users to log in frequently to obtain points, levels, and experience points.

For example, the revenue-generating behaviors of medium-frequency products such as Taobao and Tmall are mainly concentrated in ARPU (Average Revenue Per User), so the system should focus on the registration-browsing-ordering-repeat purchase route.

Another example is low-frequency products such as Qunar and Ctrip, which need to increase user stickiness by adding indicators such as the activity of content forums based on e-commerce products.

example:

(Simply list a few ways to earn points)

Coupon, membership mechanism

These are common in e-commerce and transaction products, such as Tmall, Ctrip and other product types. As a powerful tool for user operations to increase the average order value of platform users, coupons are an essential skill that operators must master. It is like Cheng Yaojin's "three axes", which type of users use low-value coupons, which type of users use high-value coupons, which type of users use long-term coupons, which type of users use short-term coupons, which type of users use all categories, and which type of users use vertical category coupons.

These can be written into a separate article, so I won’t go into details here.

(Coupon)

Data Platform

User operation is the role that is most sensitive to data requirements in operations, because the Internet era has made the data we can accept extremely rich, and also provided us with ample room to play. Here are some common data platforms I have used in the industry:

Mobile user data analysis platform: Baidu Mobile Statistics, Talkingdata Web page: Baidu Statistics, Google Analytics, GrowingIO Competitive product monitoring: Alexa, Similarweb Traffic monitoring: Webmaster Home, CNZZ (formerly Umeng+) Market monitoring: Some common index platforms such as: Baidu Index, Taobao Index can be implemented List the usage of some common platforms:

(Users’ origin and destination)

(User source type analysis)

With these platforms, how do these user data come from?

Usually, the data is buried at the bottom layer, and then cleaned, mined, and analyzed at the upper layer to obtain a result data that can be used for business decision-making.

(Where does the basic user behavior data come from?)

Operational Strategy

Tiered Operation

After having a user tag system, an operation strategy library for different users should be established.

What strategies should be adopted to recall users who have been lost in the short, medium and long term? What strategies should be implemented for users when their average LTV is lost? These are all ways of playing that can be achieved in a regular and standardized manner. Verification of fake volume of channel delivery

When we place ads in channels, we are often troubled by inflated traffic by channels and users. Then we can use the user tag system to label the common behaviors of some real users to distinguish "real users" from "fake users"

Experience conversion of newly registered users

Some users have never used certain features of the product, so targeted operational profiling can be implemented for such users, such as: PUSH, push notifications, in-site messages, jumps, small icon guidance, and banner displays can be used to target specific user operations and allow users to flow within the product.

Common user operation models

User Churn Early Warning Model

This model can be used in the early stage to operate the interval and monitor the effect. Later, it will be sent to product development to implement the entire process including automated alarm, operation strategy, and contact.

Then, these indicators are used as indicators of the churn user model, and their weights are adjusted according to the proportion of churned users. The model should be iterated according to the continuous changes in behavioral data, so as to predict as much as possible that a certain type of user is about to churn, and we need to take corresponding measures to retain them. Based on the retention effect, we can add the retention measures and retention effects of relevant lost users into the lost user model to form an automated early warning of "prediction-retention measures" as much as possible.

RFM Model

The RFM model was first commonly seen in the Internet e-commerce industry. This model focuses on post-implementation optimization. I have often used the RFM model in monitoring user stratification flow and fixed-point operation strategies. It is a model that I prefer to use. In the early stage, the operation department can first match a version of the interval value based on experience. After the digital dimensions of R, F, and M are determined, they can be submitted to the product manager and R&D colleagues to form an automated data export.

(RFM model)

SPSS prediction model

This type of model is often used in operations in the e-commerce industry, such as the Kohonen neural network model and linear regression equation prediction, which are usually used to predict DAU, GMV and other values ​​for the next month. If there are any abnormalities, operations can be carried out in a timely manner to analyze the data and formulate operational strategies.

SKU price clustering

This type of model is often used to stratify users according to different contribution value ranges, or to stratify products by SKU.

Association prediction model

This type of model is often used to discover some actions that are valuable for users to generate revenue. For example, users' actions of collecting and adding products to shopping carts are strongly associated with revenue-generating target actions such as placing orders. The weight index of these actions can also be further calculated.

The above model will not be elaborated in detail here, and a separate article on data analysis will be published later.

Common contact methods

Internal messages: They require development, but are the most cost-effective for user operations. Users can be operated according to their categories. SMS: It requires costs and is a common method used for short-term transitional, recall, and large-scale event marketing. Email: It is more commonly used abroad, but the usage rate of email among domestic users is relatively low, so it is not recommended. Push: After embedding the SDK in third-party platforms such as Umeng and Getui, push calls can be made to all users: This requires costs and is usually used for targeted maintenance, marketing, and recall of the platform's top revenue-generating users. AB Testing

(Use of AB in recall strategy)

This is the most commonly used method among Growthhackers. Everything in the world is inseparable from iteration. Then operators who are good at using AB, induction, review and summary will definitely be able to produce results in the long run.

Author: Cheng Han

Source: Ten Years of Growth Operation Notes