So I strongly disagreed with that finding, but two possible lines of reasoning: 1) Say you have a machine learning algorithm, where there could be two ways to

So I strongly disagreed with that finding, but two possible lines of reasoning: 1) Say you have a machine learning algorithm, where there could be two ways to improve the outcomes: get more data, or pick a better model. Google's massive dataset is a benefit/competitive moat of sorts to its search algorithm, but _in theory_ you could build a better algorithm or model. You can see some evidence of this in vertical specific searches (i.e. travel, shopping, etc.) and in DuckDuckGo's success. 2) The other refrain I hear a lot is that this additional data is not "zero-sum" - i.e. multiple apps can collect my location data or email or [insert data point here,] and I can onboard or offboard that data pretty easily. Therefore this data isn't a limited resource that can be hoarded like a true competitive moat.