Data Analysis of xiaobaixue --> what is active _ I (DAU)

I have been chatting with several people recently and everyone has their own views on being active. In addition, many people who are unfamiliar with terms of analysis due to some standard issues have mixed up a lot of information such as being active and retaining. later I found that this is a very real problem. In my opinion, none of the problems have become problems. Therefore, I would like to talk about active things here to help more developers who are engaged in game data analysis grow.

What is active? In our daily cooperation with the outside world, we often have information such as daily activity, weekly activity, and monthly activity, which seems simple, however, if you discover that you do not understand these definitions when performing statistical operations on your own, you may find it difficult to perform operations as analysts or even developers. The following describes three active definitions, usage methods, analysis methods, and precautions.

Daily active

Statistical Standards

There are many statistical standards for daily activity, including the number of daily active roles and daily active accounts in RPG. This type of game is generally divided into two statistical methods due to the role creation problem. Generally, the number of active daily accounts is commonly seen, which can be considered as the number of active daily users. Of course, many game rooms do not have such a multi-role concept. Therefore, the number of common daily active accounts is the best statistical standard.

Of course, there is also a statistical standard that is the unique identifier of the device, such as Mac. In this way, the number of daily active devices is counted, but the value is relatively small.

Definition criteria

Count the number of accounts that have logged on to the game on a daily basis. here we need to re-calculate the number.

For example, if 1000 accounts have logged on to the game for a day, and the total number of logins is 1600 (because some accounts have logged on to the game repeatedly), the number of daily active accounts for the day is 1000. Do not underestimate this explanation. problems often occur in actual operations. For example, we should add distinct when writing SQL statements to extract data:

Select count (distinct passportid) from playerlogintable

If the distinct statistics are not added, the total number of logins of all players will occur.

What can daily activity analyze?

In fact, only one day of daily activity can only perform a period-over-period or year-over-year analysis with the previous day or the same period of history. However, daily activity can play a far greater role than you think.

 

Core user Scale

In fact, the measurement of the core user scale is combined with the product cycle. In most games, the composition of daily activity can be divided into the following parts.

 

Among them, new login users have the greatest impact on daily active users. Generally, the proportion of new login users reaches 40%, which is actually one of the basis for determining the size of core game users.

From the perspective of composition, if new login users continue to convert stable old users, then the size of old active users is constantly increasing. At the same time, if the injection level of New login users remains unchanged, in this way, the number of core game users is increasing, and the percentage of new login users is decreasing. If the new Login User injection level is also increasing and constantly transforming to old users, that is to say, the size of core users is also increasing, so the percentage of new users will change steadily in a certain range.

The size of the core users mentioned above is measured by daily active users. The reason is that the daily unit is used as a measurement unit to objectively reflect the game enthusiasm of users, it exactly matches the shortest cyclical cycle of a user's game.

In our daily analysis, we can simply calculate the ratio of the relationships between new users and active users on a daily basis in a cycle to see a long-term trend, to some extent, it reflects the current growth of core users.

Some people may ask, how do you think about the function of reflux users?

In fact, the proportion of users contributing to daily active users is very low, but the contribution of this part cannot be ignored, because after the launch of various marketing methods such as major festivals and channel promotions, it will generate a great contribution to the daily activity of the game. But in general, the contribution of this part is relatively low.

After talking about this, what are the definitions of old users and reflux users? Here only the reference criteria are provided:

Back-to-Back users: log on to the game on the Statistics Day, but historical users who have not logged on to the game in the past seven days (the so-called historical users are non-new users who have logged on to the game in History)

Old active users: If rough calculation is performed, the calculation can be as follows:

Daily active users-daily new login users-daily return users

Of course, if you want to accurately measure the size of old users, you can define old users, for example:

Users logging on to the game on the daily Statistical Day, and logged on to the game again within the past seven days (note that there is no strict distinction between New login users, that is, the next day login of New login users is calculated as old users, the impact of this part on old users can be raised as needed ).

The following describes how to use dau to analyze the problem through several curves.

First, we need to obtain the Dau and dnu curves of the time segment, such:

 

In this figure, the trend of Dau and dnu is basically the same, and the impact of dnu on dau is still relatively large, but with the decrease of the subsequent fluctuations, we found that from 106 days to 280 days, the two curves showed a slow decline, but this was not enough to illustrate the problem. After careful observation, we found that the area between the two curves gradually reduced, this area is the part of the Dau that removes the dnu, that is, we can identify it as the part of the old user. This area is reduced, which means that the loss of users is aggravated and the control of active users is not proper, in addition, it may also be caused by the low retention rate of new users in a short period of time. You need to check the retention rate. We will not discuss it here.

After finding the above situation, we can use the difference value of the DAU-DNU to make a curve to analyze this problem. As shown in:

 

It can be seen that the difference is gradually falling, that is to say, the user activity is declining. This decline can be identified as a result of the low quality of users imported through channels in the future, it can also be caused by user cycle problems of the product itself. However, it is concluded that in this period, we need to urgently boost the growth of user scale. Therefore, we can see that the subsequent two corresponding pull operations have improved the scale.

In addition, let's take a look at the proportional curve of new users. As described above, the scale is basically at the level of 40%. However, it is worth noting that, when in a relatively stable period, even if there is a large scope of promotion and pull new growth, then the ratio change will not be too violent, the only reason is that, the number of old active users of the original game is declining, causing a large loss.

 

Of course, the loss of users and product stickiness can all be obtained through parsing dau from different angles. This should also be analyzed in conjunction with other data, such as the retention rate of the next day, data such as user traffic loss rate, startup times, and logon duration distribution can be found out for many fake dau users, such as 1-3 s users. Under normal network and design conditions, this data may be caused by many fake users, that is, cheating.

For another example, we can use Event Management to differentiate the impact of user growth in the promotion and non-promotion periods on dau, such as the impact of new login users in the natural growth period on Dau, determine the Dau quality and channel quality, or analyze the influence of New login users on dau during the promotion period.

If necessary, you can also determine the threshold value of loyal and active users based on user logon habits, such as logon times and logon days, to ensure the quality of Dau.

In fact, there are many hidden problems and analysis elements behind dau, which also need to be carried out according to your own business needs. Here we just provide you with an analysis idea and method. The specific problems should be analyzed based on specific requirements. However, DAU resolution is inseparable from the support of subdivided data and other data, but it is not always subdivided for data analysis. There are some factors that will not be achieved by dividing data, but also experience accumulation. For Analysis on this part, see:

Http://www.cnblogs.com/yuyang-DataAnalysis/archive/2012/02/08/2303909.html