Shulou(Shulou.com)06/01 Report--

The editor will share with you what user operation is. I hope you will get something after reading this article. Let's discuss it together.

What is user operation?

It aims to maximize the value of users, through a variety of operational means to improve activity, retention rate or payment indicators. In the user operation system, there is a classic framework called AARRR, that is, add, retain, active, spread, and make a profit (already covered in the historical article).

User layering

However, moving from active users to profitability are not two simple steps. If the user is active in opening the product, it is certain that the business model will be profitable? Excellent user operation system should be a dynamic evolution.

Evolution is a pyramid-level division of user groups, with a dependent relationship between the upper and lower levels.

First of all, the state of the user community is constantly changing. E-commerce, for example, registers, downloads, uses products, recommends, evaluates, buys and pays, cancels, uninstalls, and loses. From an operational point of view, we will guide the user to do what we want him to do (in this case, pay), which is called the core goal.

Of course, the core goal is not achieved overnight, users have to go through a series of processes.

Not all users will complete the steps as we imagine, and each step will show a funnel-shaped transformation. We regard the whole link as the evolution of the user group.

The image above is a typical bottom-up evolution that outlines the ideal behavior of the user community.

Since the user group is no longer a simple whole, operators can no longer operate rudely across the board, but need to operate according to different groups. This is called both refinement strategy and user layering.

Its greatest value to operators is the use of different strategies through layering.

New users: I hope they can download the product, the common strategy is the welfare of new users

Download users: I hope they can use the product, at this time should be guided by a novice, let him familiar.

Active users: I want to deepen their frequency of using the product, so the operators should continue to operate, solidify the user's habits, and be interested in the content of the product.

Interested users: I want them to make payment decisions, buy goods, and operators can use different promotions and marketing methods.

Paying subscriber: this is my target user, and I hope the user can maintain this state all the time.

Different user levels take different measures. Operation will also be limited by resources, when we can only invest limited resources, we often choose the core group, that is, the paying users above. Because according to the 2008 rule, only the core group can contribute the greatest value.

A typical example is that in game companies, there will be special manual customer service and even telephone lines to serve RMB players with a sweet voice. The average player may be an automatic reply that remains the same for thousands of years.

We must have known about layering, so how should we divide it?

In fact, there is no fixed way of layering, we can only set up a system according to local conditions according to the product form. But it has a central idea: divide it according to indicators, because indicators are a clearly measurable standard that is far better than the empirical intuition of operators.

The image above shows a simplified layering of game users, with each layer of metrics being quantifiable. In order to make the upper and lower levels of users clear, the groups should be as independent as possible, that is, when calculating RMB players, Tuhao players should be excluded, and when calculating ordinary players, the upper two layers included in the results should be excluded, so that the operation is highly targeted.

Then operators can build hierarchical reports according to this, through data trends, develop a variety of ways to improve data.

Next, let's think about the form of user layering in Zhihu. Its core is the production content of Big V? Or do more users participate in Live to generate revenue? It's hard to decide. in fact, in many operating systems, user layering is a two-tier structure.

It aims at two complementary cores to form a double pyramid stratification.

Under this structure, its core users are not only big V in content production direction, but also loyal fans in consumption direction, and they represent two types of operation strategies:

Content production direction: early use of the invitation system to obtain talents from various industries, maintain relationships through operators, and encourage production content. The mechanism of the product will also encourage the better creation and production of Big V.

The direction of content consumption is to find out the content interests of ordinary users, guide them, and cultivate their payment habits. Increase the exposure of Live, value and e-books, and design all kinds of coupons to promote users' use.

This kind of double pyramid structure aggregates content producers and content consumers together to form a virtuous circle of the whole platform: big V creates content, attracts ordinary people, ordinary people pay for content, and Big V gets benefits.

User layering in a double pyramid structure is not uncommon. Take the e-commerce that we are familiar with as an example, there are both buyers and sellers. The buyer's mode of operation is already familiar, but what about the seller? Shop opening tutorials, seller universities, shop decoration, exposure display, shop backstage, all kinds of auxiliary products … Operation also needs to help sellers grow, so sellers can also be divided into ordinary sellers, senior sellers, big customers, and super financiers.

Is O2O a double-layer structure? Certainly. Online is a user, offline is a variety of offline or service entities, but these sellers are more sales push and marketing staff maintenance, but we can also use hierarchical thinking to operate. Others are online celebrities and people who broadcast live videos, big Vs and grassroots users of Weibo, companies and employees recruiting APP, and so on.

The shape of different products will be different, and different users can be used at different stages of the same product. In the early days of a product, the goal of user layering is more users and KOL, and later, it will be closer to the business direction, which requires flexible layering for the operation.

User layering, generally four or five-tier structure is fine, too much layering will become complex, not suitable for the implementation of operational policies.

User clustering

Is there only user layering in the user operation system? Not exactly.

The user hierarchy is the upper and lower structure, but the user group can not be completely summarized by the structure. Think about it briefly, we delineate the group of paying users according to whether to pay or not, but there are also differences in this group, some users spend a lot of money, some users buy at high frequency, and some users once bought but do not buy now, how should this be subdivided?

If you continue to increase the number of layers, the conditions will become complex and will not solve the business needs.

Therefore, we use a horizontal structure of user clustering. The groups in the same layer will continue to be divided to meet the higher needs of refinement.

How to understand user clustering, let's take the following case to illustrate.

There will be significant differences between men and women in products with consumption as the core, which is two different groups. The core goal of clustering is to improve operational effectiveness and maximize the value of operational strategies. In e-commerce products, it is normal to distinguish between men and women, but in tool APP, it may not be necessary.

This is what I have always stressed, layering and clustering, are based on product and operational objectives to build a system.

Next is the practical application of clustering.

RFM model is a classic method in customer management, which is used to measure the value and profit-making ability of consumer users. It is a typical clustering.

It relies on the three core indicators of charging: the amount of consumption, the frequency of consumption and the time of the last consumption, in order to build a consumption model.

Consumption amount Monetary: consumption amount is the golden index of marketing. The 2008 rule points out that 80% of the income of enterprises comes from 20% of users, and this index directly reflects the contribution of users to corporate profits.

Consumption frequency Frequency: consumption frequency is the number of times users buy within a limited period, and the users who buy most often have higher loyalty.

Last consumption time Recency: measure the loss of users. The closer the consumption time is to the current user, the easier it is to maintain a relationship with them. The value of users who spent a year ago is certainly not as good as that of users who spend only one month.

Through these three indicators, we can easily build a coordinate system to describe the level of user consumption, and form a data cube with three indicators:

In the coordinate system, the two ends of the three axes represent the consumption level from low to high, and users will fall into the coordinate system according to their consumption level. When there is enough user data, we can divide it into about eight user groups.

For example, if a user performs well in terms of consumption amount, consumption frequency and the most recent consumption time, then he is an important value user.

If the last consumption of an important value user is relatively long ago and no longer consumes, it becomes important to retain the user. Because he used to be very valuable, we don't want to lose users, so operators and marketers can call back specifically for this group of people.

Different quadrant areas in the picture correspond to different consumer groups. Do you want to operate as a whole, or do you want to be treated differently according to the crowd?

This is the RFM model, which has been frequently used in traditional industries, but can be transplanted to be used by us in the consumption-oriented operation system. It is not only the core of CRM system, but also the core of consumer user clustering.

There are two main ways of clustering in RFM model.

One is to establish indicators, with indicators as the basis for division, which is similar to user layering.

The judgment and establishment of indicators need the experience of business experts: what is high consumption frequency, what is low, how much consumption is valuable, these are all knowledge. And it needs to be constantly revised and improved.

The figure above is a simplified division, and the practical application will be more complex because the indicators may not be representative. Most of the fee-related data will be distributed with a long tail. 80% of users are concentrated in the range of low frequency and low amount, while 20% of users generate most of the revenue, which is the difficulty of division.

Indicators are generally divided by the quantile of descriptive statistics, such as the median, the first quartile, the third quartile and so on.

The other is to use the algorithm to establish user clustering through data mining, without the need for manual division. The most common algorithm is called KMeans clustering algorithm, the core idea is "birds of a feather flock together, people are divided into groups".

We use the data of a company on the Internet for Python modeling, first deal with dimensionless (z-score), and clean out the abnormal extreme value.

The three columns of data in the figure above are standardized user consumption data. The closer the value is to 0, the closer to the average. Because the r value is the last consumption time, the smaller the value is, the closer the time is, and the higher the value is, the longer the consumption is.

Through the three indicators of RFM (called features in machine learning), we first establish a visual scatter chart. The picture below is a scatter chart of the latest charge R and the amount of charge M. Each point represents a user's fee-related data.

For the time being, the law of user clustering can not be seen on the scatter chart, and we can only make a preliminary judgment, and most of the data are centralized.

Since the core idea of KMeans algorithm is "birds of a feather flock together, people are divided into groups", it takes distance as the objective function. In short, the closer the two users are, the more likely they are to be similar, so KMeans finds similar groups and calls them clusters. The greater the distance between clusters, the more independent the user groups, which is called clustering; the more compact the distance within the cluster, the more similar users are, which is called clustering.

To speak through a chart:

The users marked in red circles are more likely to be similar and belong to the same user group. Because they in R and M these two indicators, the data are similar, are in the low consumption amount, and the recent consumption of people.

As for whether or not, let the algorithm solve it, the specific principle and process of the algorithm will not be demonstrated. Let's assume that we can divide into five user groups and then take a look at what these groups look like.

The different colors in the image above are the user groups calculated by the algorithm.

Red user group: represents the high consumption amount, because the quantity is rare, so there is no obvious distinction in the time of the last consumption, but not for a long time. These are the dads and financiers of the products.

Green user group: represents the users who tend to lose, the amount of consumption of these users is not too much, and the operation can take appropriate recovery strategies.

Purple user group: represents the recent consumption, consumption amount of less users, the operation needs to tap their value, to develop and cultivate.

Cyan and blue seem to be indistinguishable. What if we change the dimension of the scatter chart?

After switching to indicators R and F, it is a different perspective. The cyan user group has more consumption times than the blue user group, and the consumption frequency of blue users is relatively poor, which needs more encouragement. Purple users have a very high consumption frequency.

At this point, the user group has been clearly distinguished, can you accurately summarize the characteristics of these users? Although from the data distribution, the long tail shape will affect the readability to some extent, but the operation can still make the corresponding operation means for different groups.

Observe the final result through the scatter plot matrix (the picture may not be clear):

That's what the RFM model is all about. It can dynamically provide the consumption profile of users and provide a basis for fine operation for the market, sales, products and operators. This is also one of the applications of data mining in user operation, which we should understand.

How to divide groups is a subject of knowledge, the division of groups is less, the degree of differentiation is not obvious; divided more, there is no business value, how do you operate more than 20 groups? The number of groups is to strike a balance between data and business.

In a word, the method of clustering is to manually divide the user group through indicators and attributes. The other is to give business meaning to the results through data mining. Anyway, the ultimate goal is to improve operational effectiveness and value.

We can use the RFM model to try to broaden our thinking a little bit, can we do something new? You can try.

Finance: investment amount, investment frequency, last investment time

Live broadcast: how long to watch the live broadcast, the last time it was watched, and the amount of reward

Content: number of comments, number of words, number of likes

Website: number of logins, duration of login, last login time

Game: level, game duration, game recharge amount.

These are the references I have briefly enumerated, which may not be accurate and serve as a reference for all of you. The clustering strategies of different products are also different, such as hotel products, accommodation is not a solid demand, do you need to add a time dimension? Maybe the accommodation will be better grouped.

It should be noted that the number of groups is not fixed, can be two, can also be four, depending on business needs, mainly can include the majority of users. Just not too much, on the one hand, it is complex, on the other hand, the performance of KMeans clustering in multi-features is not good.

Through user layering and user clustering, we must have understood the cornerstone of the user operation system. User layering is based on the division of the general direction, what core goal do you want users to work towards, and user clustering is to improve the effect by dividing them into finer granularity. The two complement each other.

If the user is large to a certain order of magnitude, layering and clustering may not be a good method, because with the further expansion of the product, no matter how subdivided, it is difficult to meet the complexity of users, which is common in all kinds of platform products. At this time, the user profile (UserProfile) system needs to be introduced, and the layering and clustering of users are only part of the portrait.

After reading this article, I believe you have a certain understanding of "what is user operation". If you want to know more about it, you are welcome to follow the industry information channel. Thank you for reading!

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.