So I am currently working on a project for "sales people". Essentially, the user is giving us a plethora of information related to deals they are trying to win. We have data related to the price of the deal, margin of the deal, number of people working on the deal, caliber of the people working on the deal, and the risk of the deal. I want to assign a "score" to these deals based on the given metrics and visualize how the score is calculated in a simple way.

It would be very easy to visualize this all as a radar (spider) chart (I'm aware that a traditional radar chart has limited power since the ordering of metrics is arbitrary and the area formed by the spider chart doesn't hold much value) - but why can't I take the results from a correlation matrix or regression output to weight the metrics in some way and/or oder them based on importance?

An example - let's say I've built a radar chart for NBA players based on Points per game, assists per game, player efficiency, etc. If I had a logistic regression output to determine the MVP, I would use the output to order and weight the variables (based on p-values?) to build the radar chart. If I then took statistics from the most recent NBA season and plotted each NBA player, wouldn't the player with the biggest area be one of the most valuable?

I can't seem to find much information on building weighted radar charts - although I have seen that percentiles can be used for the actual metrics. Does anybody have any advice or alternative solutions?


1 Answer 1


You basically want to display the importance of each feature in the model as it relates to your final score. Rather than starting with the visualization, I would figure out the data first and then see which visualization fits. Here, they use a simple bar chart:



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.