[chrismiller]

The post was written by a 3rd party and not Instagram. The last section of the article explains how they managed to get the data:

"Four things allowed us to conduct this analysis: Instagram’s new API, their use of incrementing ID numbers to represent users and photos, the central limit theorem"

[dreamux]

Unfortunately, this isn't very rigorous - the dataset can easily be gamed by instagram.

[ajays]

If the Instagram guys have read this http://en.wikipedia.org/wiki/German_tank_problem , they might be having a good laugh...

[naithemilkman]

Well my point was more that Instagram exposed enough data such that RJ Metrics were able to get access to enough information to extrapolate the cohort retention rates.