Anonymous
I have used quite a few methods to detect outliers in my previous projects. For example, in the last project where I was trying to understand customer behaviour on ordering. I used the inter-quartile distance as a means to flag outliers specifically ones that had an order frequency 3 time beyond the inter-quartile distance. I then capped these vales at the 2nd and 98th percentile respectively. This was quite an efficient and reliable method to deal with outliers in the preliminary analysis,