A little about mobile app stores and numbers that can not be trusted
Do you think that to know the volume of the audience of your mobile application, just download it in the app store and look at the beautiful graphics? Not quite, if not otherwise.
In this article, we will analyze the data on downloads submitted by various app stores and compare them with the actual number of installations.
')
Each application store provides its own installation statistics, and often additional information. But in the modern world this data is not enough.
We released several applications and only after that we began to realize that we were doing something wrong. But to understand exactly what was wrong was quite difficult. We saw only bare digits of installations from various sites that did not give any information about our users.
To understand your customers, you need to get as much information as possible about their actions in the application. And for this there are various services analytics. We used Flurry - with its help we track the first launch of the application and determine which store the build was obtained from. Then we collect the general numbers and get the launch coefficient for the installation ( Launch / Install ), which will tell you exactly which store is most effective in this regard.
What did we see after data collection based on several different applications? Quite an interesting picture:
If in the case of Google Play, the situation is not surprising - it seems reasonable that not all users who installed the application start it at least once - then in the case of Amazon and Opera there are some anomalies.
Amazon developer console in most cases understates the real number of installations - we have this happened with all applications, except for one. And Opera in our case shows a consistently low number of real launches.
What conclusions can be drawn from this?
The first option is unlikely, since Flurry is one of the largest analytics tools and has proven itself to be quite reliable. In addition, in this case, the anomalies would not be repeatable.
The second option seems the most likely, and possibly in combination with the third. But bots that download applications “just like that,” rather than on demand, also look weird.
It is worth noting that Opera is a kind of “multiplier” and also places sent projects in a large number of other stores: Beeline , Blackberry , Nokia , Yandex.Store and others. It is likely that part of the data is transmitted incorrectly at the level of interaction with them.
The last option, associated with the audience, seems unrealistic, but it can happen - Opera users often forget what they just installed, and Amazon fans bought a couple of devices and play testers in their spare time.
In addition to the conversion of installations into launches, the expected number of installations must be taken into account. All because a platform that gives a high conversion and a low number of installations can be worse than inefficient on conversion, but with a huge number of installations.
Take the average figures from our practice and round them up to the understandable perception of the total thousand installations. More installations bring Opera and Google Play. It would seem that nothing will change further and we have a clear leader.
But here we remember our recently obtained coefficients and we understand that Opera, leading in installations by almost 30%, in fact will receive 17% fewer launches than Google Play, which is second in terms of installations. And Amazon in this and in another case is not to be taken into account at all.
The conclusion from this can be made simple - you should not completely trust the numbers in the app stores, use analytics, know your users and keep abreast of your apps.
In the future, we plan to talk more about analytics in mobile games and some of its aspects, also related to the specifics of various app stores.
We will be glad if you decide to express an opinion or share your experience on this issue.
Write in the comments if you encountered anomalies and features that came to light when analyzing installations on various platforms and markets.
Thanks for attention!
In this article, we will analyze the data on downloads submitted by various app stores and compare them with the actual number of installations.
')
Each application store provides its own installation statistics, and often additional information. But in the modern world this data is not enough.
We released several applications and only after that we began to realize that we were doing something wrong. But to understand exactly what was wrong was quite difficult. We saw only bare digits of installations from various sites that did not give any information about our users.
To understand your customers, you need to get as much information as possible about their actions in the application. And for this there are various services analytics. We used Flurry - with its help we track the first launch of the application and determine which store the build was obtained from. Then we collect the general numbers and get the launch coefficient for the installation ( Launch / Install ), which will tell you exactly which store is most effective in this regard.
What did we see after data collection based on several different applications? Quite an interesting picture:
If in the case of Google Play, the situation is not surprising - it seems reasonable that not all users who installed the application start it at least once - then in the case of Amazon and Opera there are some anomalies.
Amazon developer console in most cases understates the real number of installations - we have this happened with all applications, except for one. And Opera in our case shows a consistently low number of real launches.
What conclusions can be drawn from this?
- The analytics we use is incorrect
- These stores are partially incorrect
- Among the users of stores there are bots
- The specificity of the audience of different stores affects the number of application launches.
The first option is unlikely, since Flurry is one of the largest analytics tools and has proven itself to be quite reliable. In addition, in this case, the anomalies would not be repeatable.
The second option seems the most likely, and possibly in combination with the third. But bots that download applications “just like that,” rather than on demand, also look weird.
It is worth noting that Opera is a kind of “multiplier” and also places sent projects in a large number of other stores: Beeline , Blackberry , Nokia , Yandex.Store and others. It is likely that part of the data is transmitted incorrectly at the level of interaction with them.
The last option, associated with the audience, seems unrealistic, but it can happen - Opera users often forget what they just installed, and Amazon fans bought a couple of devices and play testers in their spare time.
In addition to the conversion of installations into launches, the expected number of installations must be taken into account. All because a platform that gives a high conversion and a low number of installations can be worse than inefficient on conversion, but with a huge number of installations.
Take the average figures from our practice and round them up to the understandable perception of the total thousand installations. More installations bring Opera and Google Play. It would seem that nothing will change further and we have a clear leader.
But here we remember our recently obtained coefficients and we understand that Opera, leading in installations by almost 30%, in fact will receive 17% fewer launches than Google Play, which is second in terms of installations. And Amazon in this and in another case is not to be taken into account at all.
The conclusion from this can be made simple - you should not completely trust the numbers in the app stores, use analytics, know your users and keep abreast of your apps.
In the future, we plan to talk more about analytics in mobile games and some of its aspects, also related to the specifics of various app stores.
We will be glad if you decide to express an opinion or share your experience on this issue.
Write in the comments if you encountered anomalies and features that came to light when analyzing installations on various platforms and markets.
Thanks for attention!
Source: https://habr.com/ru/post/299246/
All Articles