We are designing the information displays for portfolio managers. In the process of understanding what information is needed we have identified from our resident portfolio managers and customers the need to get a quick overall status of the portfolios under management.

If we are looking at one portfolio there are some specific characteristics we would like to roll up and summarize. Let’s take the example in the fixed income world. If we have a portfolio of 100 securities, when we look at that portfolio summary we would like to get a sense of the spread, duration, convexity, and credit quality. When asked how they would roll this data up, the portfolio managers tell us to use a weighted average.

The problem is that the distribution of the data is non-normal most of the time so taking an average is useless. What would you do to express the overall distribution in non-normal data?

Some thoughts:

create a plot –  it is non-normal and the distribution is irregular so perhaps a distribution plot would be useful (a little sparkline of the distribution to give a general idea)
monitor control boundariesupper and lower bounds with counts below, in the middle and above would be helpful (this would allow the user to see the outliers quickly)
measure driftpositive and negative drift from a benchmark or model where you want to be at for each position and identify the drift (this may be normal but I am not sure)

It seems to me that we could come up with many much more useful and truthful representation of the underlying data than the weighted average.

Have you played around with this?
Is there an industry standard for expressing this data?
Do you have some ideas?