in app purchase

[CMU App Store Study]

I've been reading your paper INFERRING APP DEMAND FROM PUBLICLY AVAILABLE DATA for a couple of days because I am willing to apply your methods to a research I am working on. First of all, congratulations to you and to Mr. Telang, I appreciate your work. It was exactly what I was looking for, since I need to estimate app demand in my country and I found it difficult to get data from developers. 

 

In my empirical data, I see strong influence of in-app purchase in the Top Grossing rankings. My Rank correlations (Top Paid and Top Grossing) are very weak. There is a impressive presence of Free apps and Paid with IAP in the Top Grossing rankings. Therefore, I should adjust the method to address IAP. I could not understand exactly how you estimated the value of θ. I kindly ask: could you please clarify its calculation?

 

MC, Brazil

(3/11/2013)

[Rajiv Garg]

I am glad you found our research work useful. It is always good to hear from practitioners in the real world to see how they use the academic research.

 

As you mentioned IAP option is increasingly used in the app world because that is one way app developers can hope to gain more revenue after the first purchase/download of an app. The model suggested in the discussion section is as follows:

Pdg = dp * (p+ θ*IAP)

ð  log(pdg) = log(dp) + log(p+ θ*IAP)

ð  Log(rg) = b0 + b1*log(rp) + b2*log(p+ θ*IAP)

 Which is very similar to the original model (equations 4 and 5). The difference is estimation process – you need a non-linear regression approach to solve for the unknowns whereas in the previous model without IAP we used truncated OLS model.

 

RG