I am attempting to create a portfolio optimization based on my own criteria. Instead of doing a mean-variance optimization, I am trying to optimize based on something different (let's say I am trying to target low P/E versus the variance in Earnings Growth, so presumably optimizing P/E for variance in earnings growth rate). I have a few questions, if anyone feels they can help me please: - If I swap the equations in the typical mean-variance optimization formula, I presume that I insert earnings growth rate in the place of returns. Along the same lines, I also presume that I should create a covariance matrix of historical earnings growth rates based on the data. Is this the correct approach? - I will take earnings growth rate estimates from analysts but I will estimate earnings growth variance myself. Can I just input my estimates for the variance or do I need to use another model if variance is not measured from the "return" data (aka the historical earnings growth). - Assuming I'm even framing the problem correctly, how do I then optimize the results based on low P/E ratios? I feel like Mean Variance optimization was built to encompass valuation because valuation is implied by price returns. My situation is different. Thanks in advance for your help.