We pay them money but I have to say.. there are aspects to the model which freak me out -Like the relative paucity of good R hooks, the abortive attempt at an interface via dplyr basically died. Right now, its curl. Its http sucking and no real way past pagination.

Or, how hard it can be to intuit the json you need, as a naieve user. Back porting from kibana is silly.

Or how easily you can deploy a naieve cluster which hangs bigtime on large data. This thing is like PostGres and ZFS: you really can't fake it, you have to understand it.

I see good potential in Elastic training