This comment is no longer valid as meta.objectType is added for all export types in the RequestHandler class.

I wonder whether a percentile based aggregation would not provide better insight into the questions we may have regarding output size (es docs).

Would it be more useful to know that the average output size is, e.g., 500kb or that 99% of outputs are below 700kb?

Percentile based aggregation can be more expensive to calcuate, which might disqualify it, but from what I can tell it looks pretty safe (also considering we are working with report documents so not sure there will be millions).

Something like:

output_size: {
  25th: ...,
  50th: ...,
  75th: ...,
  95th: ...,
  99th: ..., // outliers
}

Let me know what you think @tsullivan !

++ I am in favor of this. At the time I implemented it with avg, I was hoping to use median instead, but there isn't an actual median aggregation. With percentiles, we get the median from the 50th percentile. This is good!