How would you balance yield, product quality and impurity removal during process optimization?

Why

Quality attributes are not negotiable in the way yield is. ICH Q8(R2) and Q11 frame development around a target product profile and critical quality attributes, with the process designed to deliver them consistently. So the logical order is to define the acceptable ranges for attributes such as aggregates, charge variants, glycosylation and host cell protein, then search for conditions that maximize yield inside them.

Interpretation goes wrong when steps are judged in isolation. A feed change that raises titer by 30 percent may also raise aggregates and host cell proteins. If polishing must then cut deeper to remove aggregates, step yield falls, and overall yield may not improve at all.

A wider elution cut on ion exchange can raise step yield while letting more acidic variants through. The useful metric is overall yield of in-specification drug substance per batch, together with the margin to each quality limit. Another trap is reading averages.

A condition that gives the best mean yield but variable aggregate levels may produce occasional failing batches, which are far more costly than a few percent of yield. Design of experiments across linked steps, with quality responses measured alongside yield, shows these trade-offs. Small-scale data may not capture everything at scale, so margins should be kept.