Low budget user growth model!

"Growth hacking" must be familiar to those in the Internet circle, especially in the past few years. Many people have also seen classic cases from abroad. For example, Netflix analyzed the movies and programs watched by customers and found that movies starring Kevin Spacey and political TV series were very popular with users, so Netflix produced the TV series "House of Cards"; when Facebook was doing grayscale testing, it was found that the new version would reduce the monetization rate by 25%, so it urgently terminated the launch of the new version, etc. Its core concept is to rely on technology and data-driven to achieve the goal of growth.

But in recent years, everyone has discovered that "growth" is no longer attractive. The so-called "growth" is all other people's "growth" or cases from foreign countries. When they come to China, they will become unsuitable. After all, there are no operations or channel positions abroad, right? Just look at the commercialization level comparison between APP store and domestic Android application stores such as Huawei, Xiaomi, vivo, and oppo. You can see that domestic growth can be said to be a hard growth, and sometimes even the old Silicon Valley is far behind.

Another reason is that the company is not that rich and cannot support the cost of your trial and error in terms of manpower and material resources, so the boss simply tells you not to do anything fancy and just do what others do. If you are caught off guard, you will be out of your depth!

So, how can we achieve “localized” growth with a limited budget? Today I will bring you a lean activation model. First, let me explain why it is called the Lean Activation Model. The word “Lean” comes from the MVP (Minimum Viable Product) model in The Lean Startup, which refers to the minimum viable product model. It means that with little start-up capital or budget, the product is adjusted to the optimal level through experiments to achieve growth. The word “activation” comes from the AARRR Pirate Model. The model uses five English abbreviations to represent the five life cycle nodes of users, namely acquiring users, activating users, improving retention, generating revenue, and spreading recommendations. Our model is established at the user activation stage. Finally, the combination of "lean" and "activation" is the lean activation model we are going to introduce today.

Lean Activation Model

1. What is activation?

Drucker once said: "If you can't quantify it, you can't manage it." So before talking about the model, we must first understand what activation is and the quantitative standards for activation. Activation comes from the second letter A (Activation) of AARRR. It is also translated into "increasing activity rate" in China, but I prefer the original translation "activation" because activation does not refer to a single indicator such as one-sided improvement of new user retention or improvement of registration rate.

This involves two misunderstandings in the activation stage. The first is that registration is equal to activation, and once a mobile phone number is left after registration, the user is considered to be real and valid. The second misunderstanding is to only look at the retention of new users, thinking that this indicator can reflect the user activation status. However, people often overlook an important indicator, which is the utilization rate of core functions.

Activate 2 major misunderstandings

The core function is the aha moment often mentioned in "Growth Hacker". It is about how to let users use the core function of the product in the shortest time so that users will be impressed and remember your product. Sometimes, it is because the user fails to leave a deep impression on the other party during the "first meeting" that activation fails, resulting in user loss.

Different types of products have different core functions. Take the game Honor of Kings as an example. Each game has its own unique rules, and the cost of understanding is very high. It is particularly important to use novice guidance to let new players understand basic operations. Therefore, the completion rate of novice guidance, the length of the first game and the number of games become important indicators of activation.

Taking the Douyin APP as an example, the core function is to collect favorite shoes, and then you can see the price fluctuations and market conditions. Therefore, the collection rate and tool usage rate (dressing, shoe VR, etc.) of new users are important indicators of activation. Taking Yiche APP as an example, its core function is the car tool, through which you can check the lowest price and related information of your car. Then the inquiry rate of new users after using the tool becomes the activated North Star indicator.

North Star Indicator

Judging user activation is not limited to registration rate and retention rate. It is also necessary to find the usage rate of core functions as a monitoring indicator based on the product type. To sum up, registration rate, new user retention and core function usage rate become the North Star indicators of the product activation stage.

2. Reasons for activation failure

Now that we have discussed what activation is, let’s talk about the reasons for activation failure. Many users come to the APP through promotion channels, but never use it again after launching it once. This is considered activation failure. There are two main reasons for this:

(1) Channel issues: There are quality issues with the users brought by the promotion channels.

(2) Product issues: Users are not exposed to core features (aha moment) and are lost midway.

For channel issues, please refer to the previous article "How to use less money to bring better quality volume?" | Channel Quality Evaluation Model", which will explain to you how to evaluate the quality of users brought by the channel, improve user quality and thus improve activation quality. I will not go into details here. So today we are going to talk about how to solve problems from a product perspective and achieve user growth through the lean activation model.

3. Model Overview

The basic principle of the lean activation model is to conduct experimental tests at a fast pace while keeping costs under control, constantly discovering problems, proposing ideas, conducting experimental tests, and conducting review and analysis.

Model Overview

By continuously repeating these four steps, we can learn from failures and summarize successful experiences in each round of experiments, and ultimately quantitative changes will lead to qualitative changes, achieving user growth. There isn't anything fancy, it's just experiments over and over again, it's so plain and boring.

4. Model Process

In this chapter, based on the four steps mentioned in the previous chapter, we will use the Yiche APP as an example to explain how we use this model to achieve user growth.

1. Find the problem

Gideon's Law states: Writing down the problem clearly is half the solution. So before starting modeling, try to get your team to raise more questions. After a series of brainstorming, we can write down the questions and summarize them into the following problem board:

Issue Board

Since we need to increase user growth in the activation stage, what is the definition of activation? What is the current reason for the activation failure? What kind of growth model should be chosen? What is the formula for growth?

The problem has been clearly written out, so the first step of finding the problem is successfully completed, and the next three steps are to solve the problem.

2. Propose ideas

Previously, we raised four questions, namely the definition of activation, the reasons for activation failure, how to select the growth model, and how to determine the growth formula. In this chapter, we will break down these four questions one by one.

Definition of activation: User activation is not just registration, but also user retention and core function usage. Taking Yiche APP as an example, we choose to force users to register. Why do we force registration? After AB testing, it was found that the user quality of the experimental group (mandatory registration) was significantly higher than that of the control group (non-mandatory registration). Therefore, the observation and monitoring indicators chosen are the next-day retention rate and inquiry rate of new users.

Reasons for activation failure: channel problems and product problems. If the channel cannot be changed, consider how to adjust the product logic, enhance user experience, and improve activation.

Growth model: Lean activation model, which achieves growth through experimental testing.

Growth formula: In the early stage of the cold start of the experiment, we first analyze the user situation of the existing functions. Note that the users here refer to all users, including new and old users, rather than the behavior of new users. Why use all users? There are two reasons. The first is that if only new users are selected, the number is relatively small and not enough to illustrate the problem. The second is that new users are not familiar with the APP and cannot access some functions with deeper entrances, which will also affect the judgment. To summarize, the growth formula is determined by analyzing the retention and inquiry behavior of all existing users after using the function.

The full user behavior analysis is as follows:

Growth Formula

Retention growth formula:

Growth formula = Retention rate difference x penetration rate = (reached retention - unreached retention) x (reached number/DAU)

Inquiry growth formula:

Growth formula = Inquiry rate difference x penetration rate = (reached inquiries - unreached inquiries) x (reached number of people/DAU)

The retention rate difference is the retention rate of users who have used (reached) the feature minus the retention rate of users who have not used the feature. This indicator is used to determine whether this feature is an aha moment, that is, users are impressed after using it, and thus remember the product and improve retention. The penetration rate is the ratio of users who access this function to DAU. For example, if there are 2 million users who have read the article and the DAU on that day is 5 million, then the penetration rate is 40%. The permeability is used to determine whether the entrance depth of the current function is reasonable or whether it is buried too deep.

It is not difficult to see from the formula that in order to achieve growth, we need to try from three directions: retention rate difference, inquiry rate difference and penetration rate.

(1) Improving the difference in retention rate: If the difference is negative, it means that the user experience of using this function is very poor, which will cause user loss. In this case, we need to find a way to adjust or remove the page. In addition, for functions with a positive difference, can we expand the difference by adjusting the product form? For example, we can improve the quality of articles, reduce low-quality machine-written and clickbait articles, thereby improving the reader experience and improving the retention rate.

(2) Inquiry rate difference improvement: The logic is the same as the retention difference improvement, so I will not elaborate on it.

(3) Penetration improvement: When we find a function with a high retention difference but low penetration rate, we should realize that although this function is good, it may not be noticed by users because the entrance is too deep. At this time, we should adjust the entrance depth to allow more users to use the function.

Now that we have introduced the growth formula and growth method, let's analyze the specific situation and make a growth matrix (Boston matrix) chart based on the data in the growth formula table, as shown below:

Growth Matrix

Boston not only has lobsters, but the matrix diagram is also very nice! The horizontal axis of the matrix represents the retention difference, the vertical axis represents the inquiry rate difference, the diameter of the bubble represents the penetration rate, and the functions are divided into four quadrants by red crosses, which makes it easier for us to understand the situation and make decisions.

In this section, we defined user activation, analyzed the reasons for activation failure, and determined the growth model and formula based on existing user behavior. This series of operations is to prepare for the experimental test (4-3). Everything is ready, except for the test!

3. Experimental test

In this section we have to do two things. The first is to prioritize the ideas. We have a lot of ideas, but that doesn’t mean we have the ability to test all of them, so we have to pick out ideas with a greater likelihood of growth for testing.

Growth Matrix (Cold Start Group Selection)

How to select ideas with a higher probability of growth? We can select the functions in the first quadrant of the growth matrix, including sales ranking, quick car selection and price reduction ranking, and test them in the form of buckets. Then we have the following test ideas:

Test idea diagram

First, you need to select channels. Note that you do not need to conduct rough testing on all channels, but rather test them under controllable costs. For example, choose Huawei channels instead of all Android app markets. This reflects the core concept of "lean". Then, we will bucket the users and display different landing pages after the users start the app for the first time. These are the sales list, quick car selection, price reduction list and the original page (control group). 25% of the new users are allocated to each of the four pages. As for how to divide the buckets, you can refer to the AB testing platform on the market. This aspect has been very complete and can allow N solutions to interact simultaneously. To sum up, the following experimental flow chart is obtained:

Experimental flow chart

The above picture is a simple process concept diagram. We only conducted an experiment with 4 pages to give you an example. In actual business, there will be N different channels, N landing pages, and N detailed change plans, so there will be N*N*N different plans.

In short, you need to slow down in the early stages of the experiment, choose delivery channels while costs are controllable, conduct experiments with a small number of groups, and after the process of the experimental model is running smoothly, gradually speed up and make the model run at high speed.

4. Review and analysis

In the review stage, we will get 4 kinds of information, as shown below:

We know what we know: When we came up with the idea in Chapter 2, we knew that the metrics we needed to monitor were the retention rate difference and the inquiry rate difference.

We know what we don’t know: In Chapter 2, we analyzed the behavior of all users (including new and old users) to infer the pages (functional points) that may be added during the activation phase of new users. However, we don’t know whether old users can represent the performance of new users, and we need to test it.

We don’t know what we know: our intuition tells us that the Boston Matrix, a visual representation, helps us understand the business and make decisions. Of course you can choose bar charts, pie charts, etc., but our experience and intuition make me think of the Boston Matrix first, which is data sensitivity.

We don’t know what we don’t know: In the exploration and analysis area, we don’t know which page or functional point is the best, which page and what entry depth can ensure the optimal user retention rate and inquiry rate. We need to review and analyze, collect information from each review and analysis, retain positive results, replace negative results, and quickly iterate experiments to achieve user growth.

In this section, we will focus on the information that we don’t know we don’t know, that is, the review and analysis of the exploration area. Through the experimental test in the previous chapter, we obtained the following feedback data:

Data replay chart

From the perspective of the grouped performance of newly activated users, the next-day retention rate and inquiry rate of the sales list were 5.0% and 7.6% higher than those of the control group, respectively; the retention rate of the quick car selection list was 1% lower than that of the control group, and the inquiry rate was 4.7% higher; the retention rate of the price reduction list decreased by 0.4%, and the inquiry rate increased by 6.3%.

Compared with the Boston Matrix (growth matrix) using all users during cold start, the performance of the sales list, quick car selection and price reduction list is consistent with our conjecture. The original assumption is that the car selection tool is our core function, so the inquiry rate of users after contact will be higher than that of the control group. However, the retention rates of quick car selection and price reduction list are lower than those of the control group. This result is beyond our expectation. Why does this result appear? Let's take a look at the page details and find the reason.

Page details

These three pages are all core functional pages, with only some details that are different. I marked the different points in the figure to try to explain the reasons for the data differences:

Hypothesis 1: Car selection button: We can see that this button is available on both the sales list and the price reduction list. By using the car icon, it not only reduces the user's understanding cost, but also quickly classifies the cars for easy use. This may be one of the reasons for the increase in the inquiry rate.

Hypothesis 2, sales ranking: People are interested in the sales TOP ranking, which is one of the factors affecting the second retention and inquiry rate.

Hypothesis three, sales figures: actual sales figures can gain user trust and are one of the factors that affect retention and inquiry rates.

Hypothesis 4: The lowest price inquiry button is one of the factors that affect the inquiry rate.

Hypothesis five, price reduction amount: people are not interested in the price reduction amount, and car models with large price reductions are not what users want, which may be one of the reasons for the decline in retention.

Hypothesis six, price reduction list: People are not interested in the price reduction TOP list. The same as hypothesis five, this may be one of the reasons for the decline in retention.

In addition to the above speculations, we have many other speculations. For example, if this is the case with Huawei channels, is it also the case with vivo and oppo? What would it look like if the Ask for Lowest Price button turned blue? Is it possible that our control group has a low conversion rate due to problems with the recommendation algorithm?

Although they are just guesses and may be right or wrong, these guesses are valuable information and can be transformed into questions for our next round of experiments, starting a new experimental cycle and beginning from the end.

In summary, the conclusion we reached during the review phase is that these three pages are in line with our initial conjecture and are core functions. Among them, the sales ranking has the best effect and can ensure that both retention rate and inquiry rate indicators grow positively at the same time. We recommend retaining this path as the activation path for all users of Huawei channels. At the same time, the hypotheses we proposed will be tested as a seed group for the next round of experiments.

5. Experimental Summary

The above is our entire experimental process. We build an experimental model through the four steps of discovering problems, proposing ideas, experimental testing, and reviewing and analyzing. We continuously test within a controllable cost range to understand the true meaning of growth. This is the essence of the lean activation model.

We will sort out the entire model process and get the following figure:

Model flow chart

This model can also be applied to all nodes in the user's entire life cycle, including acquisition, activation, retention, monetization and dissemination stages; it can also be used in business aspects such as activities and article content. As long as you have an experimental growth mentality, anything can achieve growth through the four steps of discovering problems, proposing ideas, experimental testing and review analysis.

This is the end of the lean growth model I introduced to you today. I hope that after learning it, you can run the model quickly and trigger user growth. We also hope that it can help the vast number of Internet practitioners, so that you know all the methods and can grow without getting embarrassed.

Author: Jiang Di

Source: Jiang Di

Related reading:

How to implement a user growth plan from 0 to 100?

A must-know method for user growth: retention curve