Lei Feng network: According to the author of this article, Tan Runyang, GrowingIO data analyst. The original text was posted on the GrowingIO Technology Blog .
This article will discuss topics from what is an active user, what is retained, why to do retention analysis, the relationship between activity, retention and product growth, and how to do retention analysis.
| What are active users and retention?
In the Internet industry, we usually introduce customers by pulling new products. However, after a period of time, some customers may be gradually lost. Those who stay behind or who frequently visit our company website/app are called retention .
For a certain period of time, a user who has acted arbitrarily on a website/app, etc., is called an active user of this website/app. This arbitrary behavior may be visiting a website, opening an application, and the like.
Nowadays, we often use the so-called “daily live†(daily active user amount, DAU), “week live†(weekly active user amount, WAU) to monitor our website, and sometimes we will see our “day live†in are gradually increased over a period of time, I thought it was a very good sign, but if you do not retained analysis of the case, the result is likely to be a mistake.
For example, a company has done a lot of new activities, people have brought a lot, the number of active users is rising, but on behalf of the customer is growing? Perhaps this is just a matter of too many new people to cover up the high turnover rate. In fact, the retention of customers is gradually reduced.
The following takes everyone a deep understanding of the relationship between growth and retention.
First, apparently increasing number of active people
The horizontal axis represents time, the vertical axis represents the total number of active weeks, and the outermost curve represents the continuous increase in the total number of people, especially in the initial period. Look at every color in the graph, and it becomes fine with time. These different colors represent users who come in each week. They slowly go through with the passage of time. At the end, only a small group of people survive. The outermost curve represents the total number of people retained.
This is actually a retained stacking chart, which has accumulated the retention of each day to form a weekly active user; at the same time, there are inflows and outs, and the total is the weekly active amount.
Second, real growth in the number of active
As time goes by, we can see the different levels of color, and after a period of time becomes a smooth line, which shows that some of our customers have survived. At this time, not only did we have new users from La Simin, but also old users who had survived before, and the increase in the number of these users is the real increase in users.
Analyzing the two curves above, we can see that users are growing on the surface. However, by looking at the nature of the phenomenon, the user retention of the first image is continuously decreasing, while the second user remains stable in the later period, so the total number of users is constantly increasing. To achieve sustained and true growth, we must try to keep users.
| Why focus on retention?In the promotion of product channels, it is often required to invest in the cost of money. In recent years, channel promotion prices have been getting higher and higher, CPC/CPD/CPT/CPA and so on have continued to increase in price, and CPT prices in application stores have turned in half a year. Several times. In addition, especially for SaaS companies, obtaining a customer is very costly both in terms of time and money. It may take two to three months to acquire a customer.
Take the above left-hand side diagram as an example. Just starting this customer, we spent more than 6000 US dollars to get this customer. In the future, under normal circumstances, the customer may pay for us in accordance with a certain cash flow. For example, it pays 500 US dollars.
In this way, you will find that the upfront costs are very high. Perhaps we can only recover the costs when customers use the products for one year or two years. If this customer is lost before, losing it means that our products have lost money, and even the original has not returned.
Let's look at the picture on the right. This picture shows the cash flow receivable from each customer's cost. In the first month we got this customer and we spent 6,000 knives. Then the customer pays us every month. For example, if we pay 500 knives per month, he won't be able to reach the so-called balance of payments until the 13th month. , We began to gradually make money after 14 months. If our retention is not done well, the customer will walk away after using it for two months. Then we lose the money.
Therefore, there is a very important meaning of retention. Customers use our company's products, the longer the better, the longer the cash flow or profit it brings, which is a very core meaning of retention.
If our retention is done well, customers will always use our products and will always bring us wealth.
From the above figure we can see two points: The first one is the time spent, the longer it is better to stay in our products; the second is, the higher the profit, the better. How can the higher the profit, the better? That is, I hope that my retention rate is getting higher and higher, and the area of ​​profit will be greater and greater.
| How to perform retention analysis?The active users in a certain period of time include old users and new users. Therefore, when doing active user analysis, it is necessary to consider the proportion of old and new users, such as NDAU (New Daily Users)/DAU Ratio, to see new users in the entire active user. The proportion of. New users rely on new sources of drainage, while old users need to do well, so I will focus on how to do retention analysis.
Retention analysis overviewThe following figure shows an example:
If our product's retention is now the green line at the bottom of the figure, the vertical axis is the ratio of retention, and the horizontal axis is time. After one day, the 100% of the users we pulled up left only 35%, the 7th day became 20%, and then slowly declined, reaching an approximate 10% effect after the 60th day.
This effect we can see if we can through some aspects of improvement, let it gradually improve it?
If we let the green retention line rise to the orange line and then rise to the red line, then the retention rate on the first day will be 70% higher, the retention rate on the 7th day will be more than 60%, and it will remain for 60 days and 90 days. The rate can also be as high as 60%. This means that 60% of the 100% of the people we obtained in the past through the market have been retained after 90 days. Looking at the green line at the beginning, our 90-day retention rate is 10%. If we can achieve 60% through our efforts, this will give us income from a steady flow of wealth and cash flow.
The above is the SaaS industry, but for the e-commerce and content social industries. The same is true. Facebook’s ability to achieve $60 in advertising revenue last quarter was due to the large user base.
Three parts of the retention curveToday, we will give you some ideas through some methods of retention analysis to see how we can improve our retention rate by optimizing our products.
The following is a common retention curve. I divided it into three parts: the first part is the oscillation period, the second part is the selection period, and the third part is the stationary period.
During the oscillation period, we can see that the number of people who have come to our company's website or download the APP has dropped drastically in the previous few days, from 100% to a few ten percent or less in a few days. This period is called oscillation. period.
After the oscillation period is the selection period, under normal circumstances, the customer has a preliminary understanding of our products. He began to explore our company's products to see if this product meets some of the core customer needs. If satisfied, the customer is likely to stay; if not, the customer will be gone.
After the selection period is the stationary period, the retention rate has entered a relatively stable stage.
For different periods, we should have different strategies. In general, during the oscillation period and the selection period, we should pay attention to the retention of new users. After entering the stable period, we should pay attention to the retention of product features.
| Case: Sidekick to enhance the retention of the practiceBelow, I use the actual case to talk about retention analysis. Before we explain, let's mention it, and keep the steps of analysis. Of course, the basic steps of data analysis are similar.
By way of example, we can explain that Sidekick, a SaaS company that does email enhancements, can use its personalized template to send emails to others, and it can also monitor whether or not the person receiving the email has opened the email.
The following three retention curves represent the performance of customers' retention in the first week, second week, and third week of December 2014. This is called data monitoring. The continuous decline of the retention curve is a problem found in data monitoring.
Specifically there are two problems:
1. The retention rate in the first week has dropped significantly;
2. The second week did not appear to be what we were very much hoped for - retention was flat, and we wanted the customer to stay but it was still decreasing. Lowering means that one day we will be able to pull new people will all go away.
At this time, two goals were formulated:
1. We hope that the retention in the first week will increase and will not continue to decrease. If you are a 2C company, it may be that the retention rate will increase the next day.
2. We hope that our retention curve no longer declines and enters a flat pattern.
The target's retention curve should be as follows:
Through continuous practice and analysis, the company eventually maintained a retention rate of over 20%. How did they do it?
They have discovered the problem of the drop in retention rate, but do not yet know the cause of the problem. Next, they need to explore the reasons for the data to find the drop in retention:
We first do a user grouping, and then compare the behavior of retained and lost populations. The company that specializes in mailboxes found that among the people who lost the first week, the number of emails sent on the first day was 60%, and the second time was only about 20%. That means 60% of the users used it. The product is gone.
So next we have to ask why these users go?
So based on the user feedback made a pie chart statistics, it is clear that there are two major issues:
The first 30% did not feel value: on behalf of this product does not produce value and want to uninstall it.
The second 30% do not understand the purpose of the product: I downloaded this product, but it was not what I thought it would be.
These two parts account for 60% of the users. We often say that we must listen to the voice of the users. The first thing to be solved is the problem of 60% of the people. It can be assumed that the user does not quickly discover the value of our products.
1. Cut off the use of low-frequency featuresSo the first thing they do is that since users can't quickly discover the value of our products, then we cut down some of the complex and difficult features first, and then see if the retention rate has improved, and we find that the retention rate continues to decline. There is no improvement.
In the second attempt, since users do not know what the core value of our products is, I will give them a hint and the retention rate will continue to decline. This method will not work either.
The third attempt, since users do not know how to use our products, then we do a video, in fact, many companies are doing, and the final data shows that it is not enough.
4. Just one sentenceThey have probably done more than 20 experiments before finding a feasible solution:
They wrote a word after the user downloaded and installed the product: You can use your email in your mailbox.
They found that the product was downloaded from a web page, but the user was using it on the client. The user probably didn't think so much. He felt that since he had downloaded a plugin on the website, he would use it directly on the website instead of returning to the user. Client. So they gave a hint: You now go directly to your outlook. After adding this sentence, the retention effect has changed a lot.
This is the result of the data, previously blue, and finally raised to yellow:
After upgrading the new user retention, let's look at the second question. Our goal is to shift the retention curve in the stationary phase. For this issue, it is very important to split the product according to the function and see the retention of each function. We call it product retention analysis. We need to understand the retention rates for all of the different features in this product, which retention rates are decreasing and which are increasing. And analyze the function of reducing the retention rate to find out the reasons. On the product side, we use the methodologies described above to optimize our products, which are operations-related. We do business to replace customers. Through continuous trial and error, analysis, and monitoring, they improved some of the product's functional retention rate reductions, and then optimized them to keep retention at a higher level.
Whether we are doing retention or upgrading, it is basically the same. At the beginning, we monitored every core function we focused on. If we find an anomaly, we begin to analyze the problem. We can:
Conduct various data explorations to discover problems.
Setting a goal, for example, hopes to increase the retention curve by 10 points in half a year (10 points is very powerful).
Sidekick spent half a year from December to May. Continuous hypothesis, verification, analysis, and observation during the six months allowed for significant improvements in retention.
| Exploration: Magic Number of Your ProductTake a social app as an example.
We all know that the core reason why users can stay is whether the functional design of products can meet the core needs of customers.
If we can meet, we can not go further, whether our product design can meet customer's core requirements better, faster, and more convenient. This is the second point.
So we need to understand what features new users have used, or what behavior has happened and they stay. Further, we need to know that some of the behaviors and frequency of user visits on the website/app may leave the user for a long time and become a loyal user. Finding these behaviors and occurrences, optimizing products, and promoting users to use these features may result in higher retention rates.
We hope that when new users use the App, they will be able to say “aha!†to our products as soon as possible, hoping they can quickly discover the value of the product and stay. So I hope to find the magic number of our APP. At this stage, we are most concerned about the initial retention of users, so we need to understand the relationship between users in the early use of APP (the first week to do) and the next week, and find those with high retention behavior.
To this end, we made the following points:
1. Clearly Measured Objectives: For us, we are concerned with the relationship between the first week's retention and the next week's retention. Specifically, we want to find out which of the user’s high retention activities for the next week.
2. Determine the behavior of new users when they were on On Boarding. For example, the number of logins, the number of messages sent, the number of followers, the number of shares, the number of likes, etc.
3. Calculate the correlation between these behaviors and the next week's retention within a certain period of time, and find out the relationship between the number of different behaviors in the first week and the retention rate in the next week.
The following is a kanban of the Magic Number of GrowingIO :
We found that the following four behaviors and frequencies have a strong positive correlation with the following week for the first visit of new users within 7 days: 6 messages were sent, 8 were liked, 5 shared, and 12 were followed.
Then according to the company's strategy at the current stage, the proportion of each behavior and the degree of difficulty that can be achieved, we will use "share 5 times" and "concern 12 people" as our candidate list for Magic Number.
Then we use the A/B test to make the two groups of users reach as many of these two indicators as possible through product and operational changes. Then, we verified the two groups of users. We found that the group of users that promoted the sharing of users did not receive much improvement in the following week's retention. The group of users who focused on 12 people saw a significant increase in retention in the following week.
Finally, we will focus on 12 people as our Magic number within 7 days and use this indicator as our most important reference for measuring the effectiveness of user On Boarding.
When choosing a product that is decisive for the user's On Boarding process, it is necessary to follow this principle: all that has been done are left and all left are done; those who haven’t done haven’t left, and none left to do.
Note: The process of discovering a company's "magic number" may be far more complex than the above case, but once you find a company's magic number it will give your company tremendous growth.
| Summary
Our first case combines the user's grouping and comparison, plus the user's research, combines the questions, finds the root cause of the problem, and finally finds the product's solution.
The second Case is more focused on finding the correlation between Behavior Cohort and retention for related behaviors. And find a series of Magic Number Candidates. Then go to verify and find the company's most suitable Magic Number to drive user retention. These two cut-in angles are actually in line with the user retention methodology I explained above. With the right analytical methodologies and constant exploration, we can find ways to drive user growth.
Once the retention curve has been raised, we can do the user's realisation or even other things, so that our users will only grow slowly and finally achieve an increase in retention. We continue to pull new ones, let users continue to grow, and the remaining users slowly build up slowly afterwards. These are our most important users and can be realized. And for those users who are not stable, we must also make changes, operations, or market operations for various products, gradually allow them to become retained users, and then begin realizing the real growth of active users.
Lei Feng Network (search "Lei Feng Net" public concern) Note: This article by the author of the sandal Yang, GrowingIO data analyst. He holds a master's degree from the University of California, San Diego and served as an analyst for Emas Pro and Kyocera in the United States. He has extensive data analysis techniques and experience in case studies. Joining GrowingIO after returning home to help customers build data models and boost business growth. The original text was posted on the GrowingIO technology blog and was authorized by Lei Feng. Reprint please contact us for authorization, and keep the source and the author, not to delete the content.