I did suggest you carryout a sensitivity Optimization analysis on the constraint variables, this will allow for better understanding of the interactions of the experiemental factors on your response output. So in this case you will be able to make up for predictions of local Maxima or minima of the variables and an optimum results.

Double Pigeon Optimization using Eureqa software is highly recommended. Also Multi objective constraint Optimization with predictions will go along way to cut down cost.

Dear Miranda Procter,

I suggest you to see link and attached file on subject.

https://www.bioz.com/result/rneasy%20mini%20kit%20protocol/product/Qiagen

Best regards