How to increase user growth? A complete system architecture for user growth!

In the world of the Internet , everything is for growth. A flash of innovation may make a product successful, but it will never last long. In the field of user growth , how to reuse a set of frameworks, find a path of best practices, and add a little luck to achieve business success is a topic I have been exploring. This article and the next few articles will systematically explain the best path to user growth.

There are already many complete explanations of the definition of user growth in the market, so I will not repeat them here. In short, the fundamental purpose of user growth is to increase the number of effective users of a product over a period of time (explained in detail later), thereby increasing current and future GMV and profits to achieve commercial success.

To achieve this goal, we break down this concept into the following parts from a strategic and tactical perspective:

strategy:

• Taking advantage of the situation: The design of the product itself is in line with the development trend of the times (if you still want to start a business selling pagers in 2018, I’m sorry, even God can’t help you), it meets the real needs of users (for example, think about whether the business model of "supervising college students to memorize words and charging fees" can be paid by users in the long run), and find the best PMF ( Product Market Fit).
• Ming Dao: Do ​​the right thing at the right time with the right method (for example, when should you increase your efforts in Baidu promotion , what keywords should you buy, and what kind of population portrait should you target?).
• Optimal techniques: Clarify goals and build the most useful tools to facilitate rapid growth (How can we quickly build and continuously iterate data products and marketing tools to improve operational efficiency and reduce operational accident rates? What kind of data products have the ability to help us conduct fast and effective data analysis , and which marketing tools can help us quickly implement operational strategies with the least effort?).

Tactics:

• Know what is happening and why it is happening: build an analysis system;
• Open source is important, but interception is more important: Methodology for improving retention (number of effective users);
• To do one's work well, one must first sharpen his tools: building growth tools;
• Don’t put all your eggs in one basket: build an operational knowledge base (crowd tags, channel system, decision conversion , creative center).

1. Construction of the analysis system

The purpose of building an analysis system:

Clearly understand the current status of product development, identify problems and potential, summarize TODO items, and reasonably judge necessity and priority.

In short, the fundamental purpose of analysis is to see the business more clearly and focus resources and energy on solving the most important issues.

The focus of this sentence is on the second half. If you only do it to satisfy your curiosity or pursue analysis complexity and workload, without being able to judge which things should be done first, which things should be done later, and which things are unnecessary to do, then this analysis has no value and is just a waste of time and energy.

Build an analysis system based on the user life cycle dimension

The process of how a product's user pool is formed and how users who use our product services go from birth to death is basically like this:

If the user pool of the product is like a reservoir, we hope that users will stay in our pool as much as possible. In fact, what we do every day is basically centered around the following goals (due to limited space, these specific methods will be detailed in subsequent articles, so stay tuned):

How to establish an analysis system based on indicators

Step 1: Define new users, effective users, silent users, and lost users

The definition of user classification should be based on judgment of business experience (for example, whether a user is considered lost if he/she has not logged in for 10 consecutive days) and the company's strategic goals (order volume-oriented, gross profit-oriented or GMV-oriented).

• New users: refers to users who have just come into contact with the product and completed the entire product experience for the first time.
• Silent users: refers to users who have used a product and recognized its services, but have become less active because some of their needs have shifted to other products.
• Lost users: refers to users who were once active on the product but were hurt by a certain experience or whose needs were transferred to other competing products.
• Effective users: refers to users who can continuously contribute positive value to the company (users that the company really hopes to obtain).

The understanding varies in different industries and products. For example, the effective users of products such as Weibo, Twitter , and Instagram refer to users who are continuously active on the platform. They can be quantified by relevant indicators such as the average number of users who stay for more than 30 minutes per day, the average number of users who post at least one feed per day, and the average number of users who collect or forward at least one feed per day.

For example, products like Taobao and Meituan Waimai can be measured by the number of users who have completed at least three orders in the past seven days, the number of users who have logged in more than five times in the past 30 days and the number of users who have collected more than one item in the past seven days.

In short, the number of effective users can be calculated based on the experience of the business in which you are working, and you can determine one (or several, forming a composite indicator) that can truly and reasonably measure the number of users who continuously contribute positive value to the product to quantify the number of effective users of the product.

Taking Tmall as an example, we can define these four types of users as follows:

Step 2: Break down the core indicators

In refined operations, user attributes will be segmented into more detailed dimensions:

• New users can be divided into different levels according to their potential to measure how likely they are to remain on the platform and the value they will contribute to the product in the future. Subsequent initiatives can prioritize resources for new users with greater potential to accelerate their growth.
• Effective users can be broken down and stratified based on dimensions such as user loyalty , user stickiness, and user quality. Targeted operation strategies can be used to prioritize cultivating high-quality user stickiness for users at different levels, thereby improving user loyalty.
• Silent users are stratified according to dimensions such as user quality and activation probability, helping us find the users who are easiest to activate and have the highest value, and then give priority to activating these users.
• Lost users are stratified according to the probability of recovery and user value, similar to silent users, and we prioritize reaching and stimulating the users with the greatest chance and whom we most want to recover.

Step 3: Build migration paths for different types of users

Due to the revision of our products, changes in the market and the upgrading of user needs, users at each level may experience changes in user behavior every day. Therefore, we need to build a path change system based on what we built in the second step to observe the changes in users' minds every day, the changes in users' recognition and dependence on our products, and evaluate which operational levers can be used to specifically promote which migration path.

The user migration path, taking the high-quality user class as an example, looks like the following figure:

Because there are too many combinations after our population is broken down, there are too many possible migration paths for different types of users. In order to better see and understand the flow of our users, we can observe user changes in the following ways:

Each group of people is concerned about whether they are migrating positively or negatively. Ideally, if all groups of people are migrating positively, it means that our product is monetizing well. However, if a certain group of people has more negative migration, it means that some operational levers are needed to stimulate them to prevent continued negative migration.

For example:

If it is an e-commerce product selling milk, in order to improve the user stickiness of high-quality and high-loyal users, we optimize the copy on the product details page and add copy such as "You have purchased 5 times, place an order now to get a historical calorie absorption report" to encourage users to place orders quickly and reduce the probability of users jumping out.

Since this optimization is only aimed at high-quality and high-loyalty users (by splitting the traffic to achieve personalized products for each user), after the optimization action is launched, observe the daily net positive and net negative changes of high-quality and high-loyalty users, as well as high, medium and low stickiness users. If the net positive of high-stickiness and medium-stickiness users increases significantly and the net negative decreases significantly, it means that our optimization is effective.

Daily monitoring and review special analysis

(1) Review of operational strategies

This part looks at whether the goals of the digital strategy have been achieved and where there is room for optimization based on the conversion process.

For example, if the strategic goal is to increase the number of new users, when reviewing, we need to see how many new users were ultimately achieved based on the exposure-click-registration-purchase links, and which links are the bottlenecks for improving the efficiency of this channel; if the click-through rate from exposure to click is significantly lower than the industry standard, it means that our delivery materials need to be optimized.

(The conversion path here is just an example. When we are actually using it, we will definitely need to break it down further. For example, e-commerce registration-purchase can be broken down into registration-browsing the product list page and staying for more than 5 seconds-browsing the product details page and reaching the full details page-adding to the shopping cart-clicking to buy now-...-payment is successful)

(2) Monitoring abnormalities

The main function of this part is to help us detect product accidents in real time and repair them in time. When establishing indicators, you can make them as comprehensive as possible, covering all product conversion paths. It doesn't matter if there are hundreds of indicators. Set the thresholds and only look at the ones that trigger alarms every day. You don't need to look at the ones that don't trigger alarms.

For example, our monitoring system has an indicator called " Conversion rate of clicking the button to send SMS now - completing the verification code" with a threshold of 80%. One day, this indicator suddenly dropped to 1%, indicating that there was a problem with our SMS channel or it was attacked by hackers, and the user may not have received the SMS. In this way, we can use this monitoring system to detect the accident and repair it as soon as possible to minimize the loss.

(3) Common Misunderstandings in Daily Analysis

Since the construction of the analysis system is all based on data, newcomers to data analysis may fall into the following data traps:

1) Looking at data to satisfy curiosity

All data analysis must include clarifying the analysis goals, proposing hypotheses, and verifying hypotheses. If the goals are not clear, there are no business hypotheses based on experience that make sense, and you just think "I should look at blabla data" and make requests to the BI department, it will often result in a long wait for the data to come out. After looking at it, your curiosity is satisfied, but you find that it does not seem to be of any guiding use.

Such data analysis would have no users and would be a waste of your own time and that of your BI colleagues.

2) Look for patterns in the vast amount of data

Sometimes the task assigned by the boss may be just to "analyze our users." For such a general task, students who are not clear-minded may immediately start to extract all conceivable features of global users, and then do a pivot table of the entire data, calculating various proportions, both horizontally and vertically. If there is nothing found after one-dimensional cross analysis, they will look at two-dimensional cross analysis. The results become more and more complicated, and they may get no useful conclusions even after making hundreds of tables.

A smart person only needs to do this once. Looking at the overall picture to find patterns is using tactical diligence to make up for strategic laziness. The fundamental reason for this situation is that you have not thought deeply enough about user needs.

3) Placing unrealistic assumptions on model complexity

In places that are not performance-oriented, some people like to pursue complexity in the process of doing things, as if a goal is not achieved because the method is too simple, and it can be achieved if it is complex enough and high-end enough. Of course, if it is in the scientific field, this argument is basically fine, but for those who do operations, they are result-oriented, and continuous contribution to growth is the only goal.

Some people have not completed their KPIs , nor have they proposed any complex methodologies. They have only said and done the same old things, and they are so embarrassed to collect their salaries every month that they start to pursue complex models. It is obvious that the copywriting features of the channel are not prominent enough, resulting in low efficiency in attracting new customers. It is enough to optimize the copywriting, but they insist on analyzing the similarities between the channel user portrait and the product user portrait , and then come up with a K-means clustering algorithm for mining the field. They have come up with several models just to measure individual differences to evaluate the accuracy. The model probably looks like this:

After two months of doing this, it seems as if my life has been enriched. In fact, it is really unnecessary. The essence of user growth still revolves around user needs. Data is just a tool. Don't lose sight of the important things and become a slave to data.

2. Subsequent thinking: Why do we only build an analysis system from the user dimension here, but not expand on sales, gross profit, net profit, etc.?

Avoid changes in the macro environment and user needs from affecting business decisions, confusing product values, and shifting the focus of operational strategies.

For example:

If a supermarket's sales have been declining for three consecutive months, the owner may panic, thinking that there may be something wrong with the supermarket's operating methods , and start a series of "reform measures." But if we analyze it, if the supermarket's efficiency and quality in attracting new customers have not declined, the number of effective users has not decreased, and the number of lost users has not increased, but the users who originally bought Budweiser beer have changed to buying Nongfu Spring during this period, the GMV will decline due to the decrease in average customer spending.

If we think about this question in advance, is it necessary to pay attention to the decline in average customer spending and formulate relevant measures (such as a 10% discount on beer)? Personally, I think it is unnecessary. Instead, it is likely to cause confusion in product values ​​and a shift in the focus of operational strategies, causing delays in efforts to attract new customers, innovate, and improve service quality, which were originally intended.

The reason is that our first priority in making products is to meet user needs (if users want beer, we sell them beer; if they want to drink Nongfu Spring, we sell them Nongfu Spring) rather than creating user demand ( Steve Jobs made users feel that they should buy a new iPhone once a year, which was creating user demand. Before Steve Jobs, users did not think they should change their phones once a year).

3. Conclusion: I hope you will “go through thousands of sails and return as a young man”, return to your original intention, and do only the right things

This article mainly talks about my understanding of the user growth architecture, and the first part of it, "The construction methods and ideas of the analysis system and the pitfalls I have encountered." Later, I will try to complete the remaining topics. Due to limited space and the huge differences in the industries and products of each reader, one article cannot help everyone build a detailed analysis system for their own products. Friends who are interested are welcome to leave a message for further communication.

Next issue preview:

Open source is important, but interception is more important: Methodology for improving retention (number of effective users)

When implementing actual growth, we often fall into the "growth trap", that is, we attract a lot of new users, but the retention is particularly poor, and the "floodgate" of the user pool can never be closed. In the next article, I will focus on the specific operational methods of "understanding retention" and "improving retention", and I hope it will be helpful to everyone.

The author of this article @冀肇 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 Games