Data analysis-how to calculate the analysis retention rate

Before analyzing the retention rate, the product manager should first determine whether the business goal of the product is related to the retention rate by combining the business characteristics of the product and the growth mode of the product.

User growth can be divided into the following three modes: sticky growth, virus growth and paid growth.

Sticky growth, cultivate user loyalty and pay attention to user retention. For example, Zhihu, Luo Ji Thinking and other content products or the business model of Baidu's open platform To B, the maintenance of customer relationship and the continuation of users are very important. The product business objectives under this growth model are mainly user retention rate and user update rate.

Fission-like growth, spontaneously spread through users' social networks in a short time. This growth model is common in social products such as WeChat and 2C trading products such as Pinduoduo. The initial commercial goals of this kind of products are communication cycle and user growth rate.

Paid growth: change from free to paid, and continue to pay to increase LTV；; ; Or pay directly for the service. For example, tools such as Evernote and ProcessOn, or 3C tools such as mobile phones and cameras. The commercial objectives of such products are the conversion rate paid by users and the repurchase rate of products.

The statistical method of reservation combines the following three aspects:

(1) new and active

New retention mainly analyzes the loyalty of new users to products. If new functions are added or optimized during the analysis and the user experience is improved, then the added and retained changes are a good standard to measure the value of the functional product.

Active retention and active user retention are important methods to monitor product quality/stickiness and understand the quality of a certain channel.

General active retention > active retention, because active users are more loyal than new users.

(2) Equipment and account number

Generally, the number of devices is the same as the number of accounts, but there are two special situations: the first is to log in different devices with the same account, and when the number of devices is greater than the number of accounts, for example, the same Taobao account logs in different mobile phones; The second is that the same device logs in to different accounts, and the number of accounts is greater than the number of devices. For example, the same mobile phone logs in to two WeChat accounts.

(3) The nth day and within the nth day.

The choice within the nth day and the nth day is largely related to the purpose of analysis and the nature of the product. Generally, products that are used less frequently in a long period of time, such as hotels and tourism products, can be kept within n days (be sure to get rid of them! ), but the calculation method of retention within n days is often very high, so the reference is very small. In addition, the retention within n days is mainly to analyze the loss, that is, how many devices/accounts are lost after being active once in a period of time, but with the increase of n, more and more devices/accounts will be retained within n days, that is, with the passage of time, the number of devices/accounts will increase, which is unreasonable. Therefore, it is rarely used in practical applications and is preserved within n days.

As for the time period, we generally pay attention to the retention of the next day, the third day, the seventh day and the thirtieth day. The specific time period depends on the user's habit of using the product, the nature of the product and the purpose of analyzing the data.

There are eight statistical methods to combine the definitions of the three reserved dimensions. Taking landing behavior as an example, what conditions a specific product meets can be considered as a specific definition of retention:

Note: When statistical information is retained within n days, duplicates must be deleted, otherwise the retention time may be longer than 100%.

Taking the 7-day active equipment retention rate as an example, the calculation method of retention rate is explained:

(1)7-day retention rate =? 100%, as shown by the black line, the number of active devices at 1 1 is divided by the number of active devices at 4 places.

(2)7-day retention rate =? 100%, the day of addition is day 0, and the next day is 1 day. The reason for this calculation is that the usage habits of some products are limited by time periods, such as office products, and the weekend usage rate is low. This calculation method can avoid the calculated retention rate being lower than normal. As shown by the red line, the number of active devices at 10 is divided by the number of active devices at No.4. ..

(3) The retention rate within 7 days =, among the devices active on 1 day, the number of daily active devices that are still active from the second day to the seventh day is divided by the number of devices active on 1 day.

More date choices are also possible.

If the market remains abnormal, it can be further analyzed from the perspectives of product channels, user sources, user groups and sub-functions. Of course, every link of product, operation, technology and market will have an impact on retention, which requires multi-angle analysis, so I won't start here.

This comparison can be analyzed from many angles, not limited to user groups, user sources, user behaviors, product channels, product functions, etc. More detailed retention analysis can also filter users' behaviors and then compare and analyze retention.

For example, the retention rate of a function is higher than that of the market, or the retention rate of a function is higher, then this function is of great value to the product. By adjusting the product, more users can use this function and improve the retention rate. Or we compare the retention rate of male and female users and find that the retention rate of male users is lower than that of female users, so we can analyze the reasons, optimize product strategies, adjust product promotion channels and attract female users.

Analyze the following retention diagram (also called pistol diagram). Through the data of this chart, we can see whether the overall retention of newly activated users is getting higher and higher with the optimization of products, or whether the stickiness of products is declining with the passage of time.

The retention curve of new users has two characteristics: it drops rapidly in the early stage and tends to be stable after a period of time.

In order to improve the performance of product retention curve, we can shorten the time for users to enter the platform period (activate as soon as possible), let more users enter the platform period (activate more) or optimize the product experience by finding product problems. Of course, in the end, we still need a surprise product experience, so that users can gradually form habits and enhance the indispensability of products.