Implementation of Statistical Model to Estimate Mobile Ad Revenue

I know exactly the trauma of losing all your savings in a race of making some decent earnings with it. Regardless of the type of business, it is a nerve-racking expedition, and especially for the individuals who are brave enough to bet all their chips. Yes, we all know that one should do sufficient research before investing their hard-earned money in anything, but most of the people don’t follow this recommended practice unless they face a couple of failures or challenges.

I have developed multiple mobile games and haven’t received any remarkable income from these games. So I have decided to gamble some money on paid marketing in hope to achieve the required results. Currently, I have launched a paid advertising campaign for Alien Spaceship War Aircraft, which is a free mobile game developed in Unity3D. Alien Spaceship War Aircraft is available on iTunes, Google Play andAmazon AppStores in case you are curious to check it out.

Being jobless in Australia and having gambled more money in mobile games, I have to be more careful with my savings, so I decide to make use of my data analysis skills – applying a statistical model to forecast an outcome of my uncollected risk. Firstly, I couldn’t find anything which can be applied quickly, and it leads me to make some sound decision. Consequently, I have developed my own data model, which is quite straightforward and based on the current data of Alien Spaceship War AirCraft. This model doesn’t take consideration of any external variable.

Conclusion: If I need to make $30 a day with Alien Spaceship War Aircraft, I have to spend $15,000 in 600  days (Approx 1.6 years), and have to wait for 500 days (Approx 1.3 years). It doesn’t look encouraging at all.Even my rough calculations indicate that I should not spend this much money to generate $30 dollars a day. I know I can make $30 through UBERX, but I am not ready to lose my faith in mobile games. As my calculations don’t take many other factors in consideration so it can be totally misleading, so I decide to apply a model created by an expert. After few minutes of web searching, I found an article in which  Patrick Gleeson, Ph. D., Registered Investment Advisor has explained how to calculate Ads revenue for a mobile app. Accordingly to his calculations, an application can make $ 12,240 if it is downloaded by 100,000 unique users.

Ok, so I decide to apply Dr. Patrick’s calculations to another dataset, which is extracted from my friend’s developer account. Actually he has made $663 in last one week with  1,074,812.00  impressions, please be cautious because these impressions have appeared across all of his apps – it is not for a unique app. The analysis shows that 1621 impressions are required to generate $1. According to Dr. Patrik’s calculations, it takes just 333 impressions to generate $1. Clearly, my friend should have made $3227 with  1,074,812.00  impressions according to Dr. Patrick’s model. It might be due to the fact that I am not applying Patrick’s calculations to impressions generated by a specific application.

To resolve this dilemma, I decided to apply Patrick’s calculation on a dataset for a specific app. Thanks to Techno Keet Pvt Ltd. for providing data of one of their famous games, Duck Hunting.  The number of downloads of the game  on Google Play is over 100,000, and it can be a perfect example to apply Dr. Patrick’s calculations. The game has generated $2,588 with  13,089,529 impressions so for, which means that it took 5,507 impressions to generate a single dollar. It is quite terrifying and discouraging for any investor, who wants to invest in mobile apps development because it would take a hell of a lot of money to generate this much downloads.

In conclusion, it is evident that spending lots of money in paid marketing doesn’t look sensible unless your app is able to fly by its  own ability. No doubt, it is recommended to invest 2 to 5 grand in marketing if the application is really interesting or based on a unique idea.Moreover, I am still not satisfied with above calculations, and will keep working to design a data model which can be useable, specifically, by game developers. I would appreciate any correction, suggestion, data or any other help if anyone likes to share, which would enable me to conclude my research.