How to correctly understand user retention rate?

Retention rate is the most important indicator to measure whether a product is valuable to users, and therefore it has become one of the most concerned topics for product operators. Based on his own work experience, the author of this article summarizes and shares four key points related to retention rate, hoping to be helpful to everyone.

01 What is retention rate?

In fact, many indicators in the product field lack a common definition in the industry. It’s not that there aren’t excellent definitions, they just haven’t become ubiquitous in the industry yet. Take the simplest product conversion rate as an example. How many people can tell the difference between user conversion rate and user number ratio?

For example, the repurchase rate of product A is 60%, while that of product B is only 40%. This does not necessarily mean that A is better than B. It is very likely that A is taking shortcuts in the definition of the repurchase rate to make it look better and attract investment. From this point of view, it can be said that the period we are in now is still a period of infancy of methodology.

So when we talk about retention rate, what exactly is retention rate?

Some people say that if you come today and come again on the second day, it is retention, and if you come again on the seventh day, it is 7-day retention.

This statement is consistent with the common definition of data platform, but it is not accurate enough.

Some people say that if you come today and come again within 7 days, it means you have 7-day retention.

Don't laugh, there really was a celebrity unicorn executive who asked for this definition of n-day retention, and also convinced the CTO to develop a corresponding report for daily tracking. I was there in person at the time, so I had to compromise and come up with two concepts, one called n-day retention and the other called n-day retention.

With strong critical thinking, let's first look at the business purpose behind it to see if it is reasonable:

Wanting to see the "retention" within n days means wanting to see this kind of business performance: how many people who came on day T come back from day T+1 to day T+n? Only these people who come back can be counted as my retained users. The reason why we think we should not look at the retention on the nth day is that we feel that only looking at whether a user comes on a certain day is too random. What if the user comes on the n+1th day or the n-1th day? Wouldn’t that be included in the statistics?

If you think about it carefully, there are two problems:

1. The purpose of looking at the retention within n days is to look at the churn, that is, how many people come once and then never come back, and then use the churn to get the desired retention. The starting point is reasonable, but the method is indeed inappropriate. As n increases, more and more people will survive within n days. In other words, as time goes by, more and more people will survive, which is obviously counterintuitive.
2. The reason for not looking at day n retention is that people fail to realize that retention rate represents group characteristics rather than individual characteristics . If the winning rate of a lottery ticket is 10% and 1 million people buy it, then whether I win or not, it cannot change the fact that 100,000 people will win in the end. Similarly, for the group, it makes no difference whether a single individual comes n+1 days or n-1 days, and it has no statistical impact on how many people in the group come back on the nth day. It is meaningless to talk about retention rate for a single user. You can't say that because you won the lottery, the winning rate is 100%, because the winning rate refers to a large group of people who buy lottery tickets, not just you alone - the indicator is not applicable.

The purpose of defining retention rate is to measure how much of the traffic we obtain from various channels eventually stays and becomes our loyal users. Based on this business background, the real retention rate must be for new users. When talking about "retention" to old users, we are actually talking about other businesses, such as:

1. After new users become old users, how many of them can continue to stay? This business should be called user churn, not retention.
2. After new users become regular users, how many of them come back the next day after visiting one day? This service should be called user return visit or visit frequency, and the visit frequency is strongly related to the product form. WeChat is used more than once a day, while menstrual management apps are considered good if they are used seven or eight times a month.

As for the "next day retention" of all daily active users, it is even more meaningless. This false data indicator is completely influenced by the composition of users. One part is the retention of new users on the same day, and the other part is the next day return visit (visit frequency) of old users on the same day. When more than 90% of daily active users are new users, this "next day retention" will be very low. When more than 90% are new users, this "next day retention" is strongly related to the product form, that is, the average usage frequency of loyal users.

Therefore, the scientific definition of retention rate (daily) should be: among the new users on day T, the proportion of users who become active again on day n (i.e. day T+n) accounts for the new users on day T. Google's official statement is more concise: Percentage of new users who return each day .

02 How to view retention rate

Based on the above definition, when we look at the daily retention curve, it must be a curve like this:

Through this curve, we can clearly know how many of the new users added every day will eventually remain as time goes by. And this retention curve must be able to be fitted by a power function. Let the retention rate on the nth day be Ret(n), then we must have: Ret(n)=a*n^b

I have said so many "musts" here, in fact, I want to emphasize that the retention curve is almost a standard objective law. No matter what form the product takes, there is such a curve. They may be high or low, fast or slow, but they can all be represented by a power function. Once you understand that the retention curve is an objective law , you can understand why we can use next-day retention, 7-day retention, etc. as important product indicators: because they are all point estimates of the entire retention curve.

To put it simply, it is to reduce the dimension of an entire curve into a point , so that we can observe the changes of this point every day, and thus know whether the retention of the product has improved or deteriorated . The following figure shows the change curve of next-day retention and 7-day retention. The X-axis is the date and the Y-axis is the retention rate. The business meaning is that for users newly added on day X, the corresponding n-day retention rate is Y. Obviously, if there is no such dimensionality reduction method, we have to draw a retention curve for each day. If dozens of retention curves are put together, we cannot see the trend of retention with the date.

Of course, if there is dimensionality reduction, there must be information loss. So we still need to pay attention to the observation of retention curves from multiple angles. The retention curve has two important characteristics:

1. Rapid decline in the early stage
2. After a certain period of time, it enters a stable period

If we want to improve the retention performance of the product, we need to start from these two aspects, shorten the time it takes for the user group to enter the stable period (activate as soon as possible), and allow more users to enter the stable period (activate more).

03 The significance of next-day retention

There is always a view that their business model is a very low-frequency model, and users often consume only once every few months, such as hotels and travel, so next-day retention is meaningless and there is no need to increase second-day retention.

I would like to criticize it from multiple perspectives.

1. Do you think that if users don’t come the next day, they will come the next month? Naive, no users will remember you the next day, and after a month, users will have forgotten you and will go to the next store to buy.
2. User conversion is a process rather than a point. It must be a process from cognition, recognition to subscription. The second retention represents recognition. Only when users recognize you will there be subsequent subscriptions.
3. The lower the frequency of the product, the longer the user's decision-making chain is. It often takes several months to compare prices from three different stores, and finally a few decisive minutes to complete the conversion in the transaction chain. Think about how terrible it would be if users never visited your product during these months. Therefore, for low-frequency products, secondary retention is equally important.

And because the retention curve is a statistical curve, a 15% secondary retention rate does not mean that only 20% of users stay. After a month, a total of 40% of users may stay. However, by increasing the retention rate from 15% to 20%, you can predict that the retention rate of users after one month may exceed 50%. This is also the important value of this data indicator: to measure changes in business performance in a quantitative way and to detect changes as early as possible . Note that the focus is on measuring business changes rather than measuring the business. If you don’t understand it, it is recommended that you read it carefully.

Another derivative point is that secondary retention is not limited to specific forms such as apps and mini programs. From a higher level, it should focus on the retention of product value itself. If you can continuously let users experience the core value of your product through search engines, social media, etc., you can achieve user retention. Subscribing to and continuously reading articles on Minority through RSS subscription, even if the user does not visit the Minority App, this user is a retained user of Minority and will generate actual value, such as purchasing Minority peripherals.

04 From platform retention to segmented function retention

When we talked about retention rate above, it was for the entire product or the entire platform, such as how many new users of the e-commerce app platform are retained. When the platform provides more than one product value, a variety of segmented function retention will be derived.

Take Meituan as an example. What different product values ​​does the giant Meituan provide?

1. Order takeout
2. Book a hotel
3. Book a flight
4. Taxi
5. Shared bikes
6. Grocery shopping
7. Buy medicine
8. …

As a super app, it integrates multiple services and provides diverse product values. For each service, we can see that Meituan has marketing activities for new users. For example, users who have never used the grocery shopping service can still enjoy new user discounts when they enter the grocery shopping channel. Accordingly, as long as the concept of "new users" exists, there is a retention concept corresponding to new users: how many of the new users of the grocery shopping service will stay and continue to use the grocery shopping service.

Even if it is not a super app, you can still break down retention into more detailed categories. For example, for ordinary e-commerce, you can break it down into category retention and channel retention:

• Women’s clothing category, how many new users who browse women’s clothing category every day stay and continue to browse women’s clothing
• Flash sale channel: how many of the new users of the flash sale channel stay and continue to use the channel every day?
• …

When users are retained in multiple different services, it means that users recognize the value of the product. Accordingly, the total product value generated by the entire platform for users is greater, which means that the retention of the platform will increase as the retention of segmented functions increases.

05 Conclusion

I believe that after reading this, everyone has a relatively comprehensive understanding of the basic concepts of retention. By developing the right retention metrics, we can better identify product problems. Otherwise, it is very likely that you will do a lot of things and accidentally affect the stickiness of the product, and as a result, you will see that the "retention rate" has become higher - new users are lost, and the proportion of old users is increasing, and the overall retention rate will actually improve.

Author: Guibin

Source: Guibin