It would help if you released more details about your methodology. So you are not working on a sample of the millions of apps? You are working on all of them? Your "SDK identification engine" can detect in what technology each and single one of million apps in both stores is created with? Really? How?
Do you download each binary and examine its content? Even paid ones? Large ones? Obfuscated?
What is the error margin of your methodology? Or you are 100% precise? Sorry I am honestly interested in the details, and I didnt find error margin anywhere for this study or details about methodology.