Based on one day of information (Path to file: s3://s2m-prod-eu-visits/archive/2015/02/visits_20150203.zip)

Obviously we see that there are some outliners. Lets get it proved with box and wiskey

## Warning in x[floor(d)] + x[ceiling(d)]: NAs produced by integer overflow

Now lets cut off outliners and see what the real density of the data is:

## Warning in x[floor(d)] + x[ceiling(d)]: NAs produced by integer overflow

Looks a bit better. Also we need to remove data which is less than 0. Since that is errors in tracking or overflows. Lets see how the distribution looks like and what it can be:

Histogram with normal distribution on it:

Density Plot: